Low potency toxins reveal dense interaction networks in metabolism
© Bains. 2016
Received: 4 June 2015
Accepted: 29 January 2016
Published: 20 February 2016
The chemicals of metabolism are constructed of a small set of atoms and bonds. This may be because chemical structures outside the chemical space in which life operates are incompatible with biochemistry, or because mechanisms to make or utilize such excluded structures has not evolved. In this paper I address the extent to which biochemistry is restricted to a small fraction of the chemical space of possible chemicals, a restricted subset that I call Biochemical Space. I explore evidence that this restriction is at least in part due to selection again specific structures, and suggest a mechanism by which this occurs.
Chemicals that contain structures that our outside Biochemical Space (UnBiological groups) are more likely to be toxic to a wide range of organisms, even though they have no specifically toxic groups and no obvious mechanism of toxicity. This correlation of UnBiological with toxicity is stronger for low potency (millimolar) toxins. I relate this to the observation that most chemicals interact with many biological structures at low millimolar toxicity. I hypothesise that life has to select its components not only to have a specific set of functions but also to avoid interactions with all the other components of life that might degrade their function.
The chemistry of life has to form a dense, self-consistent network of chemical structures, and cannot easily be arbitrarily extended. The toxicity of arbitrary chemicals is a reflection of the disruption to that network occasioned by trying to insert a chemical into it without also selecting all the other components to tolerate that chemical. This suggests new ways to test for the toxicity of chemicals, and that engineering organisms to make high concentrations of materials such as chemical precursors or fuels may require more substantial engineering than just of the synthetic pathways involved.
KeywordsToxicity Metabolic network Protein structure Ecotoxicology Protein binding Non-specific binding Xenobiotic
The biochemistry we observe in life on Earth is an island in the chemical space of possible biochemistry. Not all possible small organic molecules are made by life, and the chemicals making up the metabolic pathways common to life are limited to a small number of classes of chemicals – aldehydes, polyols, amines, alpha amino acids etc.. Understanding why biochemistry uses the molecules that it does is central both to engineering biochemistry to produce useful products and to understanding how terrestrial biochemistry originated. Is the restriction on the observed chemistry of life simply because life has not evolved the catalysts needed to make other molecules, because life has not found a need for them, or because there is selection against chemistry outside ‘biochemical space’?
It is plausible to suggest that life simply has not invented the means to make some classes of chemicals. We know that life makes carbon-carbon bonds using aldol condensation and not (for example) metathesis [1, 2], although metathesis enzymes can be designed in principle . There may simply not be any functional reason for making some molecules driving the evolution of the relevant enzymatic mechanisms.
There may also be limits on what biochemistry can achieve outside those imposed by catalytic mechanisms and the function of metabolites. For example, it has previously been shown that a simple measure of the degree of saturation of a molecule may be used to indicate that molecule’s toxicity, in the absence of any other structural information about the molecule, a finding that is related to the distribution of biochemicals in chemical space . It would be surprising if this were the only such constraint on the molecules of life.
the chemical space of the biochemicals that are common to life on Earth is a small subset of the chemical space possible to the chemistry of life ('Biochemistry occupies a limited chemical space').
that chemicals outside biochemical space have a higher chance of being toxic at millimolar concentrations than chemicals that fall inside biochemical space ('Mild toxicity is correlated with ‘UnBiological’ chemical characteristics' thru 'Threshold for correlations is millimolar concentration').
that a wide range of experimental data suggests that many small molecules bind to many proteins with low millimolar affinity, which provides a mechanism for the toxicity of chemicals at millimolar concentrations ('Mechanism of Ub correlation with toxicity')
that the reason for correlation of the toxicity of chemicals and their distance from biochemistry is that life has systematically evolved proteins to avoid unwanted millimolar interactions with metabolites in order to avoid poisoning itself ('Proposed mechanism of correlation of Ub with millimolar toxicity').
The results in the paper are in two parts to reflect this reasoning. The Results and discussion sections (Figs. 2, 4 and 5) describes the chemical space of life and the low level toxicity of chemicals falling outside this space. The sections on Mechanism of Ub correlation with toxicity and Proposed mechanism of correlation of Ub with millimolar toxicity (Figs. 6 and 7) provides an explanation for this effect.
These results suggest that biochemistry is more of an integrated whole than the conventional metabolic map would suggest. This has theoretical and practical implications, which I discuss briefly at the end of the paper.
Results and discussion
Biochemistry occupies a limited chemical space
I first establish that biochemical space is a relatively small subset of the possible chemical space from which metabolism could be selected. It is a commonplace that many of the components of primary metabolism “look similar to each other” (as undergraduates learning how to distinguish the α-amino acids or the sugars of the Calvin Cycle can attest). This section establishes that this apparent limitation of metabolism to a few chemical types is a real restriction in chemical space.
These rules, and some others relating to molecular stability, were implemented in the program Combimol as previously described . The chemical space of possible biochemicals includes structures usually excluded from drug design due to their sensitivity to metabolism . From the chemical space of all such molecules, all Fragments were generated as described in [7, 8]. This provides a library of Substructures with which to probe the space of actual molecules that make up metabolism.
Life’s metabolic diversity is enormous . For the purposes of this paper, I use a small subset of metabolites that are components of the central metabolic processes of all life on Earth, and pragmatically those processes that are shown on the Roche/Expasy metabolic map . All the small molecules listed in Part 1 (“Metabolic Pathways”) were used as a set of metabolites here called “core metabolism”, a collection of 611 molecules widely used by all life on Earth (some steroid hormones were not used, as they are chemically very similar and so contribute no new chemical structural types to the data set).
Why is biochemistry apparently limited to a subset of the possible chemistry that life could perform? The next three sections demonstrates that chemicals that fall outside the chemical space occupied by biochemistry are not merely unlikely to be part of a metabolism, but interfere with that metabolism in a way as to produce a toxic effect, and the further outside ‘biochemical space’ they are, the greater that antagonism.
Mild toxicity is correlated with ‘UnBiological’ chemical characteristics
In this section I introduce a measure of how different a chemical is from the chemical space of life. I show that a greater difference is correlated with low levels of non-specific toxicity. Toxicity is related to the existence of structures in the test chemical that are different from chemical structures usually found in biology.
Chemicals can be toxic for one of three broad reasons.
Toxic chemicals can be chemically reactive, such as formaldehyde or mercury compounds, and so chemically modify the components of life. Reactive toxicity depends on specific chemical functionality. The Combimol chemical generation software automatically excludes reactive moieties, and so this is not a class of toxicity probed by these studies.
Toxic chemicals can interact with a specific molecular mechanism in the organism, and so disrupt a particular biochemical function (see discussion in [12–17]). Drugs and plant secondary metabolite toxins achieve their effect in this way. This is caused by very specific chemical structures, which confer specific toxicity on molecules that contain them. Such ‘structural alerts’ were originally identified to predict mutagenicity , but have been extended to more general toxicity prediction in programs such as DEREK , TOPKAT, MULTICASE  and others . In my terms, a ‘structural alert’ is a Fragment that has a high affinity for a specific molecular target whose blockade produces a toxic effect. In agreement with this, ‘structural alert’ approaches to toxicity prediction or other structure-activity relationship methods that try to relate large structural features to biological endpoints work well for specific toxicity mechanisms, such as HERG blockade giving rise to cardiac toxicity  or electrophilic attack on DNA giving rise to carcinogenicity .
Structural alert approaches do not work well for predicting broad toxicity endpoints, such as death . A wide range of industrial chemicals have, or are claimed to have, toxicity that is not severe or life-threatening at low concentrations, and which is not obviously linked to structural alerts, but which nevertheless cause morbidity and mortality in model organisms at higher concentrations. Interactions of some of these chemicals with various receptors or enzymes is claimed, but most are simply observed to disable or kill model species without a mechanism for their toxicity being known or postulated. It is this third class of low potency, non-specific toxicity that I have probed further below.
I use a Fragment-based approach to identify the largest part of a molecule which is different from anything found in biology. Fragment-based methods of describing molecules are well known, computationally simple approaches to describing a molecule in terms of how its structure would be drawn by a chemist . Several groups have described using a fragment-based approach for molecular description and design [24–26], claiming that building drug-like molecules from chemical fragments derived from biochemicals lead to more ‘drug-like’ results.
Correlations of Ub with toxicity endpoints
Trout 24 h
Trout 96 h
Pteronarcys (24 h)
Pteronarcys (96 h)
Bluegill (24 h)
Bluegill (96 h)
Gammarus (24 h)
Gammarus (96 h)
Lemna - non-Herbicides
Lemna - Herbicides
Here the Ub measure (i.e. how much of a molecule does not match a structure found in core metabolism) is negatively correlated with the logarithm of the concentration at which a chemical has a half-maximal effect. This might be LD50 for a lethal toxicity measure or EC50 for a non-lethal measure. A negative value of the correlation means that a larger Ub is associated with a lower concentration, i.e. with a more potent toxin.
In almost all cases, for Ub 5 and Ub 6 , there is a significant negative correlation between Ub and toxic concentration. For a wide range of living systems, from isolated mammalian cells through unicellular plants and protests to multicellular plants and diverse animal species, Ub is correlated with toxicity. This correlation is highly statistically significant. The “***” level of significance in Table 1 is an indication of p < 0.000714 that the indicated correlation will be produced by chance. There were around 70 correlations performed for this initial analysis of the data (35 data sets, including two not shown sub-dividing the rat and mouse data into pharmacological categories, which had little effect, correlated with 2 Ub endpoints). If Ub was uncorrelated with toxicity, there is only a p = 0.05 chance that we would observe one “***” level correlation in this data set.
The one exception to the pattern of correlation of Ub with toxicity is Saccharomyces cereviseae, which shows only weak correlation between the toxicity endpoints reported here and Ub5 or Ub6. A possible reason for this will be discussed below in the section on thresholds for correlations.
I emphasize what this does and does not show. The correlations show robustly that molecules with segments that are not represented in the chemicals of core metabolism have a higher chance of being toxic at any given concentration level than molecules made up of structures found in core metabolism. The larger that “UnBiological” segment is, the more toxic the molecule is. None of the molecules tested for toxicity here are normal components of central metabolism (arguably with the exception of ethanol).
Correlation of Ub with other biological endpoints
Number of compounds
Whole organism toxicity endpoints
Trout (24 h)
These two data sets differ only in the time of exposure – 1 and 3 days
Trout (96 h)
Pteronarcys (24 h)
These two data sets differ only in the time of exposure – 1 and 3 days
Pteronarcys (96 h)
Bluegill (24 h)
These two data sets differ only in the time of exposure – 1 and 3 days
Bluegill (96 h)
Gammarus (24 h)
Combined data from G. fasciatus, G. lacustris and G. Pseudolimnaeus
These two data sets differ only in the time of exposure – 1 and 3 days
Gammarus (96 h)
Rodent toxicity data was manually curated from The Merck Index. Note that ‘molar’ values for mammalian whole organism studies are calculated as moles/kg body mass
Mutated colony formation
Data collected and provided by Choracle Ltd, derived from Toxnet 
Tumour formation frequency
Only compounds with at least two compound concentrations reported included
Lemna - non-Herbicides
Lemna gibba and Lemna minor
lack of growth/leaflet reduction
Compounds developed for reasons other than their herbicide effect
Lemna - Herbicides
Lemna gibba and Lemna minor
lack of growth/leaflet reduction
Compounds developed as herbicides (primarily for macroscopic land plants)
Cell numbers (combination growth inhibition and death)
Data-set heavy on chlorinated and nitrated aromatic compounds
Mostly drug-like molecules: See methods section for details of this analysis
Cell number (cell growth vs. cell killing)
Cell culture assay, not whole organism. Cytotoxicty data from the NCI anti-HIV compounds screening programme.
Ion channel blockade
Ion channel assay in cloned receptor assay, not whole organism test
Receptor binding IC50
Receptor binding assay, not a cell- or organism-based assay
Narcosis (reversible lack of motion)
Histamine receptor blockade
A variety of related structures, including anti-psychotics
Structures sets used for docking
PDB structural data sets used for docking
2e2b 1 m52 1iep 3k5v 3qri 3qrk 3g6g 1ab2 2g2h 2hiw 2gqg 2hz0 3cs9
1ald 2ald 4ald
1a94 1kj4 2bpz 2qhz 2qi6 2r5p 2r5q
1f0j 1ro6 1ro9 1ror 2qyl 3frg 3gwt 3hmv 3o57
1i7g 1kkq 2npa 2p54 2rew 2znn 3et1 3kdu
Drugs can also be toxic in the wrong place or dose. In Table 2 I also show that Ub is uncorrelated with the potency of compounds for just two drug targets (for which I had data to hand), antihistamines and NSAIDs that inhibit cycloxygenase-2. Neither show any significant correlation with Ub.
I do not claim that other toxicity or pharmacology endpoints will not be found to correlate with UnBiological. The examples in Table 2 are included to make the point that Ub is correlated with broad, whole-organism toxicity, not necessarily with target-specific mechanisms.
Reasons for variability of correlation
Table 1 provides robust statistical evidence for believing that UnBiological is correlated with whole organism toxicity. However the degree of correlation varies substantially between species, as does the statistical significance of that correlation. This could be due to genuine biological differences, or differences in the chemical space being sampled. The issue of chemical space coverage is significant. For example, an initial study suggested a strong correlation of Ub with the potency of phosphodiesterase-4b inhibitors (data not shown). However this was based on analysis of the data in two QSAR studies on PDE4b inhibition. The chemicals in the two studies were very similar to each other (i.e. were two specific series of chemicals). In effect, Ub was being used to classify compounds into the two studies, one of which was developing a much more potent drug series than the other. Therefore Ub could identify more potent PDE4b inhibitors, but for the trivial reason that it was identifying two studies looking at two classes of chemicals. When a wider set of PDE4 inhibitors was analysed, the correlation was reduced.2 It seems likely that, as with other QSAR methods, UnBiological will work best on a chemical set spread uniformly across the chemical space that is to be analysed. Bias in the molecules that happen to have been investigated to generate the data analysed here may be a cause of the differences in correlation of Ub and toxicity. This can only be addressed by collecting a more systematic set of multi-species toxicity data on defined chemicals. Data filed for the REACH legislation  may provide such a data set in the future.
Statistical significance is a function of sample size. It is not practical to collect hundreds of toxicity endpoints from all the species involved, and not desirable to discard endpoints from species that have been extensively tested. Therefore this aspect of variability has been retained in the study.
Threshold for correlations is millimolar concentration
Figure 5 shows trends in most of the larger data sets that lower potency toxins have better correlation with Ub than higher potency toxins. For the NCI cytotoxicity and Fathead Minnow data (Fig. 5d and e respectively) there is little trend for UB6, although for UB5 only the highest concentration data (−3 to 0 band) shows a statistically significant correlation (i.e. the certainty range is below 0). For all other data sets, both UB5 and UB6 show negative correlation of Ub with concentration (i.e. confidence limits are <0) only for the highest concentration band. Figure 5 confirms, for these data sets, that the correlation of UnBiological with toxicity is an effect seen primarily in compounds that have low intrinsic toxicity.
This observation may explain the failure to observe a correlation of Ub with toxicity in Saccharomyces. In the data set analysed here, Saccharomyces was tested for the effects of chemicals at six concentrations from 1.3 uM to 100 uM. Thus no chemical with an IC50 of >100 uM could be detected in this screen, and so the IC50 values analysed here are all below the threshold at which a statistically robust correlation of Ub and toxicity would be expected.
Mechanism of Ub correlation with toxicity
The observation that a simple and non-specific measure of chemical structure like Ub might be correlated with toxicity is unexpected. The observation that the correlation is more pronounced for weak toxins is, on the face of it, baffling. QSAR measures of biological potency are usually more effective for the most potent agents – whether toxins, drugs, hormones or other effectors. The findings in Figs. 4 and 5 therefore require a mechanistic explanation for the correlation of Ub with toxicity that operates at millimolar but not micromolar concentrations. This second part of the paper, and the results shown in it, address the plausibility of a potential mechanism.
The mechanism I propose here is that many, probably most chemicals will interact with some, maybe many, components of the cell at millimolar affinity.
The distinction of small molecules into ‘ligands’ and ‘non-ligands’ is a convenient classification for small molecules with respect to their effects on a specific protein, but it is a fiction not reflected in chemical reality. It suggests an absolute distinction between ‘binding sites’ and ‘non-binding’ sites. In reality, a small molecule can interact with atoms across the surface of a protein, and can often bind to proteins in more than one conformation and at more than one site (discussed further below). Only those sites which are unique and bind molecules with much higher affinity than any other site are called ‘binding sites’. The reality of the other sites that can, and do, interact weakly with small molecules is however illustrated by experimental evolution studies, where new protein functions are typically created by selecting new modes of interaction between protein and ligand from weak interactions already present in the original protein .
There is a substantial body of literature that suggests that many, maybe most small molecules can interact with many, possibly most proteins at millimolar concentration. I summarise three lines of such evidence below. This data will be very familiar to those involved in pharmaceutical screening programmes.
High-throughput screen data
High-throughput screening (HTS) is a common route to discovering novel biological function in large libraries of chemicals. In an HTS campaign, a very large library of chemicals (tens to hundreds of thousands) is tested at one or a few concentrations in an entirely automated assay designed to give a simple, semi-quantitative measure of whether a chemical interacts with a specific molecular target. Compounds that reach a threshold of activity (“hits”) are then taken on for further study. Such large screening programmes are a common approach to drug discovery .
Typical reports of such screens report a ‘hit rate’ of between 0.1 and 1 %, and report finding ‘hits’ that bind to the target protein with micromolar affinities [32, 33]. Most freely available databases of the biological effects of molecules also assume that compounds either bind to a target with micromolar affinity or better, or that they do not . Such databases imply that ‘not binding’ at the tested concentration means not binding at all. However this literature is misleading. More detailed reports of HTS campaigns routinely report widespread “non-specific” interaction of small molecules with protein or cellular targets (see e.g. [35–38]). Assay conditions, screening concentrations, detection thresholds and other factors are tuned to achieve a hit rate of <0.1 % in what in reality is a continuum of binding.
Fragment-based screening (FBS) seeks to identify small molecules that bind with relatively low affinity to proteins, and then combine these into larger molecules that bind with greater affinity (reviewed in [41–45]). FBS actively looks for high micromolar or low millimolar affinity of small molecules to proteins. It is a commonplace for researchers in this field that many small molecules (“fragments”) bind to most proteins at low millimolar concentrations. For example,  comment that “Novice users [of fragment-based screening by BiaCORE] are often surprised to see how often small molecules bind indiscriminately to proteins when compounds are assayed at high concentrations”. The data they give suggests ~75 % of a 1000 compound subset of the Maybridge Ro3 library bound equally well to two different targets at high micromolar to low millimolar concentrations. Congreve et al.  find that 90 % of compounds in their library have some binding at mM affinities. Hubbard  found a ‘hit’ rate of between 1.5 and 4.7 % when measuring small molecule binding to proteins at 0.5 mM by NMR. Giannetti  reviewed 20 different fragment-based screens, and report that all show ‘non-specific’ binding at affinities of 1 – 4 mM, although the highest affinities found ranged over three orders of magnitude in the different experiments. Spurlino  found that between 5 and 50 % of a library bound to target protein crystals at 5 mM (depending on library/target combination).
Other Non-Specific Binding observations
Non-specific interactions are a fact of life for pharmaceutical researchers, even among molecules that are selected for their specificity of action. Even in launched pharmaceuticals, supposedly selected for their singular, specific interaction with one target or target class, multi-target interactions are being recognised as the rule rather than the exception [49, 50]. LaBella commented “The non-specificity of drugs is a generally acknowledged truism” over 20 years ago , with genome-scale testing of molecules confirming that nearly all small molecules bind to multiple proteins . Such ‘non-specific effects’ now being accepted as a critical part of drugs’ actions [52, 53]. Houk et al. review a range of studies of binding of small molecules to proteins and cyclodextrin mimics of protein binding sites, and find an average binding affinity of ~0.5 mM .
Molecular mechanism of low affinity binding effects
It is worthwhile touching briefly on potential mechanisms of millimolar binding of compounds to proteins, and the likelihood that this will materially affect the protein’s function. Again, we must challenge the conventional model of a ligand binding to a ‘binding site’ on a protein. Structural studies have shown that many proteins can bind a diversity of chemical structures through adaptation of their structure (reviewed in [55–57]). Many proteins exist in dynamic equilibrium with partially or completely unfolded structures, some being dominantly disordered [58–61]. Post translational modification  or ligand binding [63, 64] can switch proteins from a disordered to a more ordered state, switches which can be related to their regulation and function [59, 65]. Proteins can also have multiple ordered, metastable structures (reviewed in ), and different folding states can be selected by ligand binding and have significantly different biological function (see for example [67–71]). There can also multiple folding paths leading to each of those states (see e.g. ).
A molecule that binds even weakly to one folding state of a protein and not to another will bias the population of protein folding states by stabilising the bound state over the others (by definition, if a small molecule M binds to folding state A, then the combination of A + M must be more stable than A and M on their own, and hence more likely to occur). If one of the structures in the spectrum of structures has a function absent from other structures, then binding of the small molecule will change that function by changing the amount of the functional conformer. The binding need not be ‘tight’, and may not even be detectable on the canonical crystal structure for the protein, but will nevertheless affect function in the cell.
In conclusion, it is found in many types of experimental systems that all, or nearly all, small molecules interact with many proteins with low millimolar affinity, and these low affinity bindings can have significant biological affect through modulation of the population of structures adopted by a protein. This observation leads both to an explanation of the mechanism of millimolar toxicity, and to its correlation with UnBiological.
Proposed mechanism of correlation of Ub with millimolar toxicity
Selection against protein binding of metabolites
The observations above that many molecules interact with many cellular targets at millimolar concentration, and that these are likely to have significant biological effects, raises an obvious question. If many molecules can interact with many proteins at low millimolar levels, and such interaction has adverse effects on the cell, and many metabolites are present in the cell at low millimolar concentration, then why does the cell not poison itself with its own metabolites?
A plausible explanation is that the proteins (and other large molecular constituents of the cell) have evolved to avoid interference from the cell’s normal constituents. A protein that needs to interact with (say) an α-amino acid will evolve a binding site for that α-amino acid. A protein that does not require interaction with an α-amino acid for its function may nevertheless have a low affinity binding site for an α-amino acid in one of its conformers by chance. If this low affinity binding site has an adverse effect on the cell, then it will be selected against. In short, any non-specific interaction of the cell’s normal constituents will be selected against just as there will be positive selection for beneficial interactions.
Thus we would expect any binding site or pocket on a protein that could bind an amino acid to be selected against unless that interaction provided a beneficial effect on the function of the protein. Any compound that ‘looked like’ an amino acid (i.e. had similar chemical groups arranged similarly in space) would therefore also not find binding sites on that protein. Similarly there would be selection against random or fortuitous binding sites for the chemical features present in sugars, lipids, phosphate esters and other common structures in metabolism. However there would be no selection against low affinity, random binding to flurocarbons, organosilicon compounds or other chemicals quite different from anything normally in a cell. These, therefore, would be free to bind to any cellular protein if, by chance, a binding site happened to exist for them. The larger the segment of the xenobiotic that was unlike the chemistry of life, the greater the potential affinity for such a non-canonical binding site. The association of UnBiological with toxicity shown in Table 1 is therefore a consequence of the failure of biochemistry to be selected to avoid random binding of chemicals that the cell does not usually encounter.
Testing the hypothesis with molecular docking
If we confine ourselves to comparing molecules of the same size and atomic constitution, then some of the artefactual results shown in Fig. 7a might be avoided. Figure 7b shows the comparison of the predicted energies of binding of α-amino acids compared to β-amino acids to five mammalian proteins, selected to represent a mix of functional classes of proteins for which multiple structures and many authentic ligands were known. α-amino acids are core metabolites in mammals, β-amino acids are not part of normal mammalian metabolism with a couple of exceptions. For some but not all proteins tested, α-amino acids are predicted to bind with lower affinity than β-amino acids, as predicted by the hypothesis. The exceptions found here are HIV protease (which would be expected to bind amino acids, as they are related to its substrate) and PPAR-γ (for which I have no explanation). Repeating this exercise with more sophisticated models that took the dynamics of proteins as well as ligands into account (e.g. [77, 78]) might produce more useful results.
Conflict with pharmaceutical experience
The suggestion that molecules that are not like biological molecules are more likely to be toxic appears paradoxical to the pharmaceutical chemist, as many drugs have potent (and hence potentially toxic) effects precisely because they are close molecular mimics of known metabolites. Thus, steroid drugs are potent precisely because they mimic natural steroids, dideoxyribonucleotides block viral DNA synthesis because of their mimicry of normal nucleosides [79, 80], penicillins mimic peptidoglycan components , and so on. However these molecules have been selected by evolution or by chemists to both mimic a specific biological effector and not to have any other effects than their target pharmacology. It is a truism of drug discovery that achieving this combination is extremely hard, and that unexpected or ‘off-target’ effects are a common cause of failure in drug discovery and development programmes [82–84]. Some of these effects are due to the close structural similarity between members of families of proteins, so that a drug selected to bind with high affinity to one target will be likely to bind to another, structurally similar target. However other ‘off-target’ effects are not obviously related to the known structural similarity of the ‘off-target’ proteins . Yamanishi et al.  suggest that this is because small regions (equivalent to the Fragments used in my analysis) confer protein binding. A substantial fraction of the effort in drug discovery programmes is tailoring the specificity of the candidate drug to bind to a small number of targets, and many launched drugs actually bind to more than one protein family [87–90]. Drugs are therefore a special case, the result of extensive selection by man to fit with biology. The same explanation is true for the observation that chemicals that are not metabolites but fall within ‘Biochemical Space’ have a higher chance of being toxic even in the absence of selected pharmacology .
Detoxification and resistance
A second apparent conflict with pharmaceutical experience is that organisms can and do tolerate a wide range of compounds that are toxic through tolerance, detoxification, and resistance mechanisms. The first two of these are less important to my general thesis than they might appear, and the third actually supports it.
Tolerance to a toxin or drug is almost invariably caused by changes in the organism’s physiology to compensate for the action of the drug or toxin. This is classically true of pharmacological agents such as alcohol, nicotine or heroin, but also to classic toxins such as arsenic. Mechanisms that oppose the effect of the drug or toxin are induced to restore a more normal physiological state. This is unrelated to the mechanism of intoxication in the first place.
Detoxification is a broad approach to removing toxins from an organism. It usually relies on enzymes (such as the CYPs in mammalian liver ) or transporters (such as the PGP family ) with very broad substrate specificities. It can also involve physical separation of the toxin into a defensive cell compartment. Compartmentalization is a common strategy for cells to sequester damaging metabolic chemistry from cell components that that chemistry might damage (e.g. oxidative phosphorylation in eukaryotes, anamox in prokaryotes). Sequestering misfolded proteins, damaged cell components or toxins can be seen as a form of ‘internal exile’, analogous to the export of these materials.
Acquisition of resistance be through one of two broad mechanisms. Detoxification mechanisms can be increased, often by mutation that increases expression of the relevant protein – this is a common mechanism of acquisition of drug resistance in cancer cells and in bacteria ([92, 93]. The other mechanism is for targets of the toxin to mutate so that they no longer bind the toxin. This is common for drug resistance [94–96]. It is not expected to apply to low potency, low molecular weight toxins, because (following my argument above) many, possibly most proteins would have to be mutated to evade toxicity. The mechanism of organisms’ resistance to chemicals other than drugs has not been reviewed systematically, so we do not know if this prediction is true.
I have shown above that molecules that contain segments that are not similar to common components of metabolism are more likely to show toxicity at millimolar levels than compounds that have chemistry similar to life. I relate this to the widespread observation that many chemicals bind to many proteins at low millimolar levels, and that this can materially affect the function of those proteins.
This observation is an explanation for the observation that the chemistry of life occupies a small corner of the chemical space. In order to function, the components of the cell must interact with each other appropriately, both with functional interactions between the macro-molecules and metabolites of the cell and with the absence of unwanted interactions. Each new chemical added to metabolism requires adaptation of the whole proteome to accommodate the new chemical. Once a complex, self-perpetuating metabolism has evolved, adding to it will be an increasingly demanding evolutionary task, not an impossible one but one that the pragmatic mechanisms of evolution will tend to avoid.
This finding has two implications. Firstly, UnBiological could be used as a measure of the chance that a new molecule is toxic. Such broad toxicity predictions are less useful than predictions of specific mechanisms of toxicity, and UnBiological specifically does not provide a mechanistic explanation. It is also only as statistical estimate. From the data analysed here, Ub 5 or Ub 6 could be used to give an order-of-magnitude estimate of the potency of a low-potency toxin, but would say nothing about high potency toxicity. It is possible that coupling Ub with other measures  might give more accurate estimates. As an initial screen for ‘drug-like’ properties [97–100], however, such a statistical indicator could find a use.
In this application of predicting toxicity, a strong limitation of the analysis presented here is that it takes no account of the concentration of metabolites in the cell. Metabolic intermediates present at nanomolar concentration are given the same weight in the Ub calculations above as common components such as glucose or glycine. One would however expect the selective pressure on proteins to avoid binding glucose to be much stronger than the pressure to avoid binding metabolites present at nanomolar concentrations.
An extension of this work would therefore include a concentration term in the calculation of UnBiological. This would include two components – consideration of the differing metabolomes of different cells or organisms, and quantitative consideration of the concentration of metabolites in an organism. In this study, a single collection of metabolites (“core metabolism”) was used to define UnBiological. I expect that predictions of toxicity based on the actual intracellular metabolome of a specific species would be more accurate for that species (and less accurate for other species) than this generic approach. This is however a substantial undertaking, involving re-calculation of most of the comparisons presented here for each species, and so has not been attempted in this paper: my goal here is to show that this approach is theoretically and practically interesting. It might also be valuable to weight the contributions of metabolies to the Ub calculation according to their intracellular concentration, although this is fraught with difficulty as intracellular concentrations of metabolites are very hard to measure, and in any case are modulated by the protein binding that this study postulates occurs promiscuously and universally. Future work could also explore the size of the overlap necessary to define Ub: again, this would be doable, but time-consuming, and so has been left for future work.
The second implication of this work is in the field of metabolic engineering and synthetic biology. Engineering an organism to produce a new chemical or execute a new metabolic pathway has been thought to require the expression of suitable enzymes to make the chemical and any intermediate or precursor molecules at sufficient concentration, efficiency, and from suitable feedstock. The rest of the cellular machinery is generally viewed as a ‘chassis’ on which to attach these changes [101–103]. For chemicals or gene products produced at low concentrations this is likely to be true . However if the goal of the engineering is to produce a chemical at substantial levels , then the analysis in this paper suggests that many aspects of the cell must be engineered, especially if the chemical to be produced is very different from one usually present in the cell.
provided a quantitative half-effect concentration estimates (i.e. not single concentration toxicities)
provided data on at least 50 compounds from diverse chemical families
were available from a small number of sources (for practical reasons, data sets of 200 compounds studied in 200 papers were not used)
were from species with recognised use in toxicity testing.
All EC50 values were as reported in the relevant papers or databases except those for Saccharomyces cereviseae, where EC50 values were calculated from the raw inhibition data downloaded from . The Saccharomyces data set was filtered to exclude organometallic compounds, to exclude mixtures or salts other than halogen or alkali metal salts, compounds for which growth inhibition at the highest concentration was <50 % or for which the growth inhibition at the lowest concentration was >50 %, and compounds for which the range of calculated EC50s across the 13 strains tested in this data set (calculated as [maximum(EC50)-minimum(EC50)]/average (EC50)) was >1. The resulting data set represented well defined organic compounds with EC50s within the experimentally measured concentrations and consistent toxicity across a range of Saccharomyces cereviseae strains.
Chemical structures were collected as. MOL files, and compiled into an SDF file for processing.
Molecules of metabolism
The chemical space of metabolism was taken as all the molecules shown on the printed version of Part A (intermediary metabolism) of the Roche/Expasy metabolic chart, with the exception of the steroid hormones. Steroid hormones were omitted because they represent many elaborations on the same core (sterol) structure, and do not add significantly to the diversity of chemical types. The chemical space of 611 metabolites is referred to as ‘core metabolism’ in this paper.
Generation of ‘all’ molecules in chemical space
The space of all possible chemical structures was explored by the program Combimol . In brief, the program generates chemical structures based on SMILES strings . The program aims to generate all the molecules of a specified size (defined as number of non-hydrogen atoms), using a subset of C, N, O, S, P or Si that is specified by the user. ‘All molecules’ are here defined as all 2-D chemical structures that have a bonding pattern consistent with the valencies of the atoms used: four bonds for C and Si, 3 for N, 2 for O, 3 or 5 for P, 2, 4 or 6 for S. For the work described in this paper, silicon was excluded from this list, and only P(V) and S(II) and S(VI) (sulphate) were used, consistent with those elements’ use in metabolism.
The SMILES language is a simple, text-based method for coding chemical structures . The program starts by generating an exhaustive list of all possible linear atom strings up to the desired size: if the maximum size was 4, it would generate CC, CCC, CCCC, CCCN, CCCO, CCCP, CCNC etc.. It then replaces single bonds with double bonds in any position allowed by the elements (CCC = C, CC = CC etc.). It then generates cyclised versions of these strings (C1CCC1, where the two ‘1’ symbols represent atoms that are connected – the reader is directed to  for a further description of the SMILES language). A molecule can have up to two ring systems, including fused rings. These form unbranched ‘core’ molecules. In a final step, the ‘core’ molecules are joined to each other to form branched molecules: thus CCC and CC could be joined to form CC(C)C.
A number of rules are included in this process to remove atom combinations that would be unstable (such as COOOC).
As described in  the program is not completely exhaustive (it under-represents quaternary carbon centres, for example): however I estimate that it generates over 90 % of the molecular structures that are consistent with the rules set provided. The program, and a new, more systematic version, which was not used in this work but will be for future work, is available for non-commercial applications from the author.
Molecular matching and similarity
A molecule is a set of atoms connected by bonds, in which all the atoms’ valencies are filled. It is assumed that all valencies that are not explicitly linked to another atom in the description of the molecule are filled with hydrogen atoms. (“molecule” here is completely consistent with the common understanding of the term, and so will not be capitalised: I define it here solely for completeness).
A Fragment is a set of atoms connected by bonds, in which the valencies of at least one of the atoms are not filled. A Fragment therefore represents part of the structure of a molecule, and not a real physical entity.
Substructure. Molecule or Fragment A was said to be a substructure of molecule or Fragment B when all of the non-hydrogen atoms and all the bonds of molecule or Fragment A could be overlaid on molecule B in at least one position.
Overlap. Molecule or Fragment A and molecule or Fragment B are said to have an N-atom overlap when the largest Fragment which is a substructure of molecule A and of molecule B has N atoms.
To identify matches and Overlaps between molecules and Fragments, I used a 2D fragment-based molecular descriptor system that has previously been described [7, 8], and proven effective in building models to predict toxicity outcomes. In summary, I generate an exhaustive set of Fragments from all the molecules used in this study as follows. For each pair of molecules, the maximum common structure (Maximum Common Subgraph – MCS) is found by ‘overlapping’ the 2D molecular structure. The MCS between each of these Fragments and between each Fragment and each original molecule is then computed to generate further Fragments. This is repeated until no new Fragments are found. The result is a list of all the molecular Fragments that are present in two or more of the molecules in the set. Fragments of 1 or 2 atoms are ignored.
Molecular descriptors of a molecule are then computed by matching a set of these molecular Fragments to that molecule, and counting the number of distinct ways that a Fragment can be exactly mapped onto a molecule. A descriptor is an integral count of the number of occurrences of a Fragment in a molecule. The molecule as a whole is described by the pattern of Fragment descriptors.
Molecular matching and molecular Fragment generation were performed by software build by Amedis Pharmaceuticals Ltd. (see [7, 8] for details) and kindly provided by Dr. Antranig Basman. Conversion of data files for transfer between programs was done with a number of small programs written specifically for this project in Qbasic, and compiled with the QB64 compiler . All programs other than MolDescrip are available from the author on request, and source code for programs other than those originating from Amedis Pharmaceuticals are also available. General chemical database manipulation was done using the CambridgeSoft ChemBioOffice suite version 12.0, under site licence to MIT. All work was done on standard PCs running various versions of Windows depending on their age.
Docking was done using AutoDock Vina . The ‘binding site’ was defined as the entire surface of the protein for all the proteins. Potential ligands were docked to a number of structures for each protein, as listed in Table 4. Ligands were docked using default parameters except for ‘Exhaustiveness’, which was set = 100. The binding energy of a ligand to a protein was taken as the maximum (most negative) binding energy of any ligand conformation to any site on any of the tested protein structures.
Calculation of ‘Unbiological (Ub)
‘Unbiological’ is a measure of the size of a region of a molecule that is not represented in metabolism. In this paper ‘metabolism’ is taken to be the set of ‘core metabolism’ molecules defined by the 611 chemicals listed in the Roche/Expasy metabolic map, as described above (section Molecules of metabolism).
What is meant by ‘not represented’ depends on the size of the Fragment that is being considered. If we only require one atom similarity between a metabolite and a test molecule, then clearly almost all molecules can ‘match’ a core metabolite. Thus Ub depends on our definition of similarity.
In this study, I define Ub as follows: Fig. 3 illustrates this process. The metabolites of ‘core metabolism’ are broken into N-atom fragments. A region is ‘unbiological’ if it does not contain an exact match to any of the fragments generated from core metabolism. Thus in Fig. 3, three metabolites (top left) and three test chemicals (top right) generate 12 Fragments, of which 9 (bottom left) completely match the original metabolites. The three Fragments not found in metabolites (bottom right) are UnBiological. These are matched to the target molecules (bottom right). The size of the largest UnBiological fragment that can be matched to a test molecule is its Ub value.
This is a measure of the size of a sub-region of a molecule that has an arrangement of atoms unlike an arrangement found in biology.
As noted in the text, this depends on the original set of biochemicals used as a definition of ‘biochemistry’. The use of central metabolic pathways is convenient, but could be improved.
Ethics and consent
This work involved no human or animal experimentation, and so no ethical or other consent was relevant.
Availability of data and materials
The complete data set of Ub values, toxicity endpoints for the chemicals analysed in this study is available for download from LabArchives (https://mynotebook.labarchives.com/), at DOI https://doi.org/10.6070/H4VQ30PJ (direct URL for download of the spreadsheet https://mynotebook.labarchives.com/share_attachment/Bains_Data/MTkuNXwxNTAzOTUvMTUtMy9UcmVlTm9kZS83NjQwNTkxNDl8NDkuNQ==). A ZIP file of the MOL files for rhe chemicals used in this study can be downloaded from https://mynotebook.labarchives.com/share_attachment/Bains_Data/MTkuNXwxNTAzOTUvMTUtNC9UcmVlTm9kZS8yMzAwNTAxMjU0fDQ5LjU=. The programs used for this specific analysis were proprietary to Amedis Pharmaceuticals Ltd (Cambridge, UK), when that company existed, and are available as compiled code only: consequently they are not available for general use. However similar chemical matching functionality can be found in RDKit, (http://www.rdkit.org/docs/index.html), which is an open-source, Python based platform.
No hits for searches for this structure in the Dictionary of Natural products.
Many thanks to Antranig Basman for making the fragment-generation and MCS-matching software available for this study, to Janusz Petkowski for many insightful comments and discussions, to the two reviewers of this article whose comments helped me make it substantially clearer, to the attendees of the Gordon Research Conference on Synthetic Biology (June 9–14 2013, Mt Snow Resort, Vermont) for support and encouragement, to Carl and Barbara Berke for their generous support and accommodation during some of this study, and to Sara Seager for her generous and unstinting financial and intellectual support.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Hoveyda AH, Schrock RR. Catalytic asymmetric olefin metathesis. Chem – A Eur J. 2001;7(5):945–50. doi:https://doi.org/10.1002/1521-3765(20010302)7%3A5%3C945%3A%3Aaid-chem945%3E3.0.co%3B2-3.View ArticleGoogle Scholar
- Grubbs RH, Chang S. Recent advances in olefin metathesis and its application in organic synthesis. Tetrahedron. 1998;54(18):4413–50. http://dx.doi.org/https://doi.org/10.1016/S0040-4020(97)10427-6.View ArticleGoogle Scholar
- Gastegger M. De-novo enzyme design for olefin metathesis. Vienna: University of Vienna; 2013.Google Scholar
- Bains W. A trip through chemical space: why life has evolved the chemistry that it has. In: Pontarotti P, editor. Evolutionary biology: genome evolution, speciation, coevolution and origin of life. Cham, Switzerland: Springer; 2014. p. 371–94.Google Scholar
- Bains W, Seager S. A Combinatorial approach to biochemical space: description and application to the redox distribution of metabolism. Astrobiology. 2012;12(3):271–81.PubMedView ArticleGoogle Scholar
- Yu B, Reynisson J. Bond stability of the “undesirable” heteroatom–heteroatom molecular moieties for high-throughput screening libraries. Eur J Med Chem. 2011;46(12):5833–7. http://dx.doi.org/https://doi.org/10.1016/j.ejmech.2011.09.044.PubMedView ArticleGoogle Scholar
- Bains W, Basman A, White C. HERG binding specificity and binding site structure: evidence from a fragment-based evolutionary computing SAR study. Prog Biophys Mol Biol. 2004;86:205–33.PubMedView ArticleGoogle Scholar
- Bains W, Gilbert R, Sviridenko L, Gascon J-M, Scoffin R, Birchall K, et al. Evolutionary computational methods to predict oral bioavailability QRPRs. Curr Opin Drug Discov Devel. 2002;5(1):44–51.PubMedGoogle Scholar
- Seager S, Bains W. The search for signs of life on exoplanets at the interface of chemistry and planetary science. Sci Adv. 2015;1:e1500047.PubMed CentralPubMedView ArticleGoogle Scholar
- Roche, ExPaSy. Biochemical Pathway Maps. (http://web.expasy.org/pathways/). 2015.
- DNP. Dictionary of Natural Products. 2015. http://dnp.chemnetbase.com/. Accessed 1/12/2015.
- Sanderson DM, Earnshaw CG. Computer prediction of possible toxic action from chemical structure; the DEREK system. Hum Exp Toxicol. 1991;10(4):261–73. doi:https://doi.org/10.1177/096032719101000405.PubMedView ArticleGoogle Scholar
- Netzeva TI, Pavan M, Worth AP. Review of (quantitative) structure–activity relationships for acute aquatic toxicity. QSAR Comb Sci. 2008;27(1):77–90. doi:https://doi.org/10.1002/qsar.200710099.View ArticleGoogle Scholar
- Ashby J, Tennant RW. Chemical structure, Salmonella mutagenicity and extent of carcinogenicity as indicators of genotoxic carcinogenesis among 222 chemicals tested in rodents by the U.S. NCI/NTP. Mutat Res/Genet Toxicol. 1988;204(1):17–115. http://dx.doi.org/https://doi.org/10.1016/0165-1218(88)90114-0.View ArticleGoogle Scholar
- Hansch C, Hoekman D, Leo A, Zhang L, Li P. The expanding role of quantitative structure-activity relationships (QSAR) in toxicology. Toxicol Lett. 1995;79(1–3):45–53. http://dx.doi.org/https://doi.org/10.1016/0378-4274(95)03356-P.PubMedView ArticleGoogle Scholar
- Schultz TW, Cronin MTD, Netzeva TI. The present status of QSAR in toxicology. J Mol Struct THEOCHEM. 2003;622(1–2):23–38. http://dx.doi.org/https://doi.org/10.1016/S0166-1280(02)00615-2.View ArticleGoogle Scholar
- Schultz TW, Cronin MTD, Walker JD, Aptula AO. Quantitative structure–activity relationships (QSARs) in toxicology: a historical perspective. J Mol Struct THEOCHEM. 2003;622(1–2):1–22. http://dx.doi.org/https://doi.org/10.1016/S0166-1280(02)00614-0.View ArticleGoogle Scholar
- Ashby J, Tennant RW. Chemical structure, Salmonella mutagenicity and extent of carcinogenicity as indicators of genotoxic carcinogenesis among 222 chemcials tested in rodents by the US NCI/NTP. Mutat Res. 1988;204:17–115.PubMedView ArticleGoogle Scholar
- Ridings JE, Barratt MD, Cary R, Earnshaw CG, Eggington CE, Ellis MK, et al. Computer prediction of possible toxic action from chemical structure: an update on the DEREK system. Toxicology. 1996;106:267–79.PubMedView ArticleGoogle Scholar
- Greene N. Computer systems for the prediction of toxicity: an update. Adv Drug Deliv Rev. 2002;54:417–31.PubMedView ArticleGoogle Scholar
- Amini A, Muggleton SH, Lodhi H, Sternberg MJE. A novel logic-based approach for quantitative toxicology prediction. J Chem Inf Model. 2007;47:998–1006.PubMedView ArticleGoogle Scholar
- Richard AM. Future of toxicology - predictive toxicology: an expanded view of 'chemical toxicity'. Chem Res Toxicol. 2006;19(10):1257–62.PubMedView ArticleGoogle Scholar
- Livingstone D. The characterization of chemical structures using molecular properties: a survey. J Chem Inf Model. 2000;40:195–209.View ArticleGoogle Scholar
- Warr WA. Fragment-based drug discovery. J Comput Aided Mol Design. 2009;23:453–8.View ArticleGoogle Scholar
- Van Voorhis WC, Hol WGJ, Myler PJ, Stewart LJ. The role of medical structural genomics in discovering new drugs for infectious diseases. PLoS Comput Biol. 2009;5(120):e1000530.PubMed CentralPubMedView ArticleGoogle Scholar
- Dobson PD, Patel Y, Kell DB. Metabolite-likeness’ as a criterion in the design and selection of pharmaceutical drug libraries. Drug Discov Today. 2009;14(1–2):31–40. http://dx.doi.org/https://doi.org/10.1016/j.drudis.2008.10.011.PubMedView ArticleGoogle Scholar
- Hancox JC, McPate MJ, El Harchi A, Zhang Y. The hERG potassium channel and hERG screening for drug-induced torsades de pointes. Pharmacol Ther. 2008;119(2):118–32. http://dx.doi.org/https://doi.org/10.1016/j.pharmthera.2008.05.009.PubMedView ArticleGoogle Scholar
- Gutendorf B, Westendorf J. Comparison of an array of in vitro assays for the assessment of the estrogenic potential of natural and synthetic estrogens, phytoestrogens and xenoestrogens. Toxicology. 2001;166(1-2):79–89. http://dx.doi.org/10.1016/S0300-483X(01)00437-1.PubMedView ArticleGoogle Scholar
- Abraham MH, Rafols C. Factors that influence tadpole narcosis. an LFER analysis. J Chem Soc Perkin Trans. 1995;2(10):1843–51.View ArticleGoogle Scholar
- Williams ES, Panko J, Paustenbach DJ. The European Union’s REACH regulation: a review of its history and requirements. Crit Rev Toxicol. 2009;39(7):553–75. doi:https://doi.org/10.1080/10408440903036056.PubMedView ArticleGoogle Scholar
- Aharoni A, Gaidukov L, Khersonsky O, Gould SM, Roodveldt C, Tawfik DS. The 'evolvability of promiscuous protein functions. Nat Genet. 2004;37(7):73–6.PubMedGoogle Scholar
- Hertzberg RP, Pope AJ. High-throughput screening: new technology for the 21st century. Curr Opin Chem Biol. 2000;4:445–51.PubMedView ArticleGoogle Scholar
- Sundberg SA. High-throughput and ultra-high-throughput screening: solution- and cell-based approaches. Curr Opin Biotechnol. 2000;11:47–53.PubMedView ArticleGoogle Scholar
- Williams AJ, Tkachenko V, Lipinski C, Tropsha A, Ekins S. Free online resources enabling crowd-sourced drug discovery. Drug Discov World. 2009;2009(Winter):33–9.Google Scholar
- Rawlins P. Current trends in label-free technologies. Drug Discov World. 2010;2010(3):17–26.Google Scholar
- Leeson PD, Springthorpe B. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat Rev Drug Discov. 2007;6:881–90.PubMedView ArticleGoogle Scholar
- Hopkins AL, Mason JS, Overington JP. Can we rationally design promiscuous drugs? Curr Opin Struct Biol. 2006;16(1):127–36. http://dx.doi.org/https://doi.org/10.1016/j.sbi.2006.01.013.PubMedView ArticleGoogle Scholar
- Ekins S. Predicting undesirable drug interactions with promiscuous proteins in silico. Drug Discov Today. 2004;9(6):276–85.PubMedView ArticleGoogle Scholar
- NCI. AIDS Antiviral Screen. 2010. https://wiki.nci.nih.gov/display/NCIDTPdata/AIDS+Antiviral+Screen+Data. Accessed 21/3/2010.
- NCI. NCI-60 Analysis Tools. 2010. http://discover.nci.nih.gov/cellminer. Accessed 21/3/2010.
- Hubbard RE, Murray JB. Experiences in fragment-based lead discovery. In: Kuo LC, editor. Methods in enzymology (vol 493): Fragment-based drug design - tools, practical approaches and examples. Amsterdam: Elsevier; 2011. p. 509–31.View ArticleGoogle Scholar
- Liszewski L. Capitalising on fragment-based drug discovery. Genet Eng News. 2009;29:8.Google Scholar
- Myrska D, Paul J. Exploring the horizons of small molecule drug discovery: the evolution and application of the ideal fragment library. Drug Discov World. 2012;Winter:51–8.Google Scholar
- Turnbull AP, Swarbrick ME. Harnessing fragment-based drug discovery at CRT. Drug Discov World. 2009;2009(Fall):57–64.Google Scholar
- Murray CW, Rees DC. The rise of fragment-based drug discovery. Nat Chem. 2009;1(3):187–92.PubMedView ArticleGoogle Scholar
- Congreve M, Rich RL, Myszka DG, Figaroa F, Siegal G, Marshall FH. Fragment screening of stabilised G-protein-coupled receptors using biophysical methods. In: Kuo LC, editor. Methods in enzymology (vol 493): Fragment-based drug design - tools, practical approaches and examples. Amsterdam: Elsevier; 2011. p. 116–36.Google Scholar
- Giannetti AM. From experimental design to validated hits: a comprehensive walk-through of fragment lead identification using surface plasmon resonance. In: Kuo LC, editor. Methods in enzymology (vol 493): Fragment-based drug design - tools, practical approaches and examples. Amsterdam: Elsevier; 2011. p. 169–218.View ArticleGoogle Scholar
- Spurlino JC. Fragment screening purely with protein crystallography. In: Kuo LC, editor. Methods in enzymology (vol 493): Fragment-based drug design - tools, practical approaches and examples. Amsterdam: Elsevier; 2011. p. 321–56.View ArticleGoogle Scholar
- Pérez-Nueno VI, Venkatraman V, Mavridis L, Ritchie DW. Detecting drug promiscuity using gaussian ensemble screening. J Chem Inf Model. 2012;52(8):1948–61. doi:https://doi.org/10.1021/ci3000979.PubMedView ArticleGoogle Scholar
- LaBella FS. Molecular basis for binding promiscuity of antagonist drugs. Biochem Pharmacol. 1991;42(supp):S1–8.PubMedView ArticleGoogle Scholar
- Makowski L, Rodi DJ. Genome-wide characterization of the binding repertoire of small molecule drugs. Hum Genomics. 2003;1(1):41–51.PubMed CentralPubMedView ArticleGoogle Scholar
- Bianchi MT. Promiscuous modulation of ion channels by anti-psychotic and anti-dementia medications. Med Hypotheses. 2009;74:297–300.PubMedView ArticleGoogle Scholar
- Campillos M, Kuhn M, Gavin A-C, Jensen LJ, Bork P. Drug target identification using side-effect similarity. Science. 2008;321(5886):263–6.PubMedView ArticleGoogle Scholar
- Houk KN, Leach AG, Kim SP, Zhang X. Binding affinities of host–guest, protein–ligand, and protein–transition-state complexes. Angew Chem Int Ed. 2003;42(40):4872–97. doi:https://doi.org/10.1002/anie.200200565.View ArticleGoogle Scholar
- Ma B, Shatsky M, Wolfson HJ, Nussinov R. Multiple diverse ligands binding at a single protein site: A matter of pre-existing populations. Protein Sci. 2002;11(2):184–97. doi:https://doi.org/10.1110/ps.21302.PubMed CentralPubMedView ArticleGoogle Scholar
- Carlson HA. Protein flexibility and drug design: how to hit a moving target. Curr Opin Chem Biol. 2002;6(4):447–52. http://dx.doi.org/https://doi.org/10.1016/S1367-5931(02)00341-1.PubMedView ArticleGoogle Scholar
- Hammel M. Validation of macromolecular flexibility in solution by small-angle X-ray scattering (SAXS). Eur Biophys J. 2012;41(10):789–99. doi:https://doi.org/10.1007/s00249-012-0820-x.PubMed CentralPubMedView ArticleGoogle Scholar
- Mittag T, Kay LE, Forman-Kay JD. Protein dynamics and conformational disorder in molecular recognition. J Mol Recog. 2010;23(2):105–16.Google Scholar
- Schultz JE, Natarajan J. Regulated unfolding: a basic principle of intraprotein signaling in modular proteins. Trends Biochem Sci. 2013;38(11):538–45. http://dx.doi.org/https://doi.org/10.1016/j.tibs.2013.08.005.PubMedView ArticleGoogle Scholar
- Forman-Kay Julie D, Mittag T. From sequence and forces to structure, function, and evolution of intrinsically disordered proteins. Structure. 2013;21(9):1492–9. http://dx.doi.org/https://doi.org/10.1016/j.str.2013.08.001.PubMed CentralPubMedView ArticleGoogle Scholar
- Uversky VN. Proteins without unique 3D structures: Biotechnological applications of intrinsically unstable/disordered proteins. Biotechnol J. 2015;10(3):356–66. doi:https://doi.org/10.1002/biot.201400374.PubMedView ArticleGoogle Scholar
- Bah A, Vernon RM, Siddiqui Z, Krzeminski M, Muhandiram R, Zhao C, et al. Folding of an intrinsically disordered protein by phosphorylation as a regulatory switch. Nature. 2015;519(7541):106–9. doi:https://doi.org/10.1038/nature13999.PubMedView ArticleGoogle Scholar
- Wright PE, Dyson HJ. Linking folding and binding. Curr Opin Struct Biol. 2009;19(1):31–8. http://dx.doi.org/https://doi.org/10.1016/j.sbi.2008.12.003.PubMed CentralPubMedView ArticleGoogle Scholar
- Uversky VN. Unusual biophysics of intrinsically disordered proteins. Biochim Biophys Acta. 2013;1834(5):932–51. http://dx.doi.org/https://doi.org/10.1016/j.bbapap.2012.12.008.PubMedView ArticleGoogle Scholar
- Mitrea DM, Kriwacki RW. Regulated unfolding of proteins in signaling. FEBS Lett. 2013;587(8):1081–8. http://dx.doi.org/https://doi.org/10.1016/j.febslet.2013.02.024.PubMedView ArticleGoogle Scholar
- Behar I, Lezon TR, Yang L-W, Eyal E. Global dynamics of proteins: Bridging between structure and function. Ann Rev Biophys. 2010;39:23–42.View ArticleGoogle Scholar
- Pinkas DM, Strop P, Brunger AT, Khosla C. Transglutaminase 2 undergoes a large conformational change upon activation. PLoS Biol. 2007;5(12):2788–96.View ArticleGoogle Scholar
- Stefl S, Nishi H, Petukh M, Panchenko AR, Alexov E. Molecular mechanisms of disease-causing missense mutations. J Mol Biol. 2013;425(21):3919–36. http://dx.doi.org/https://doi.org/10.1016/j.jmb.2013.07.014.PubMed CentralPubMedView ArticleGoogle Scholar
- Shan Y, Eastwood Michael P, Zhang X, Kim Eric T, Arkhipov A, Dror Ron O, et al. Oncogenic mutations counteract intrinsic disorder in the EGFR kinase and promote receptor dimerization. Cell. 2012;149(4):860–70. http://dx.doi.org/https://doi.org/10.1016/j.cell.2012.02.063.PubMedView ArticleGoogle Scholar
- Yang S, Blachowicz L, Makowski L, Roux B. Multidomain assembled states of Hck tyrosine kinase in solution. Proc Natl Acad Sci. 2010;107(36):15757–62. doi:https://doi.org/10.1073/pnas.1004569107.PubMed CentralPubMedView ArticleGoogle Scholar
- Francis DM, Różycki B, Koveal D, Hummer G, Page R, Peti W. Structural basis of p38α regulation by hematopoietic tyrosine phosphatase. Nat Chem Biol. 2011;7(12):916–24.PubMedView ArticleGoogle Scholar
- Lindorff-Larsen K, Piana S, Dror RO, Shaw DE. How fast-folding proteins fold. Science. 2011;334(6055):517–20. doi:https://doi.org/10.1126/science.1208351.PubMedView ArticleGoogle Scholar
- Nixon SL, Cockell CS. Nonproteinogenic D-amino acids at millimolar concentrations are a toxin for anaerobic microorganisms relevant to early earth and other anoxic planets. Astrobiology. 2015;15(3):238–46. doi:https://doi.org/10.1089/ast.2014.1252.PubMedView ArticleGoogle Scholar
- Kasam V, Salzemann J, Botha M, Dacosta A, Degliesposti G, Isea R, et al. WISDOM-II: Screening against multiple targets implicated in malaria using computational grid infrastructures. Malar J. 2009;8(1):1–16. doi:https://doi.org/10.1186/1475-2875-8-88.View ArticleGoogle Scholar
- Baxter CA, Murray CW, Waszkowycz B, Li J, Sykes RA, Bone RGA, et al. New approach to molecular docking and its application to virtual screening of chemical databases. J Chem Inf Comput Sci. 2000;40(2):254–62. doi:https://doi.org/10.1021/ci990440d.PubMedView ArticleGoogle Scholar
- Waszkowycz B, Perkins T, Baxter CA, Li J, Liebeschuetz J. Receptor-based Virtual Screening of Very Large Chemical Datasets. 2001. http://acscinf.org/docs/meetings/220nm/presentations/220nm16.pdf.
- Klepeis JL, Lindorff-Larsen K, Dror RO, Shaw DE. Long-timescale molecular dynamics simulations of protein structure and function. Curr Opin Struct Biol. 2009;19(2):120–7. http://dx.doi.org/https://doi.org/10.1016/j.sbi.2009.03.004.PubMedView ArticleGoogle Scholar
- Shaw DE, Maragakis P, Lindorff-Larsen K, Piana S, Dror RO, Eastwood MP, et al. Atomic-level characterization of the structural dynamics of proteins. Science. 2010;330(6002):341–6. doi:https://doi.org/10.1126/science.1187409.PubMedView ArticleGoogle Scholar
- De Clercq E. Trends in the development of new antiviral agents for the chemotherapy of infections caused by herpesviruses and retroviruses. Rev Med Virol. 2005;5(3):149–64.View ArticleGoogle Scholar
- Jarvis B, Faulds D. Lamivudine: a review of its therapeutic potential in chronic hepatitis B. Drugs. 1999;58(1):101–41.PubMedView ArticleGoogle Scholar
- Wanxman DJ, Strominger JL. Penicillin-binding proteins and the mechanism of action of beta-lactam antibiotics1. Ann Rev Biochem. 1983;52:825–69.View ArticleGoogle Scholar
- Entzeroth M. Emerging trends in high-throughput screening. Curr Opin Pharmacol. 2003;3:522–9.PubMedView ArticleGoogle Scholar
- Oprea TI, Marshall GR. Receptor-based prediction of binding affinities. Perspect Drug Discov Des. 1998;11:35–61.View ArticleGoogle Scholar
- Polley MJ, Burden FR, Winkler DA. Simulation and modelling of chemical and biological complex systems. Aust J Chem. 2006;59:859–64.View ArticleGoogle Scholar
- Keiser MJ, Setola V, Irwin JJ, Laggner C, Abbas AI, Hufeisen SJ, et al. Predicting new molecular targets for known drugs. Nature. 2009;462(7270):175–81.PubMed CentralPubMedView ArticleGoogle Scholar
- Yamanishi Y, Pauwels E, Saigo H, Stoven V. Extracting sets of chemical substructures and protein domains governing drug-target interactions. J Chem Inf Model. 2011;51(5):1183–94. doi:https://doi.org/10.1021/ci100476q.PubMedView ArticleGoogle Scholar
- Lu J-J, Pan W, Hu Y-J, Wang Y-T. Multi-target drugs: the trend of drug research and development. PLoS ONE. 2012;7(6):e40262. doi:https://doi.org/10.1371/journal.pone.004026.PubMed CentralPubMedView ArticleGoogle Scholar
- Cavalli A, Bolognesi ML, Minarini A, Rosini M, Tumiatti V, Recanatini M, et al. Multi-target-directed ligands to combat neurodegenerative diseases. J Med Chem. 2008;51(3):347–72.PubMedView ArticleGoogle Scholar
- Frantz S. Drug discovery: Playing dirty. Nature. 2005;437(7061):942–3.PubMedView ArticleGoogle Scholar
- Ohlson S. Designing transient binding drugs: A new concept for drug discovery. Drug Discov Today. 2008;13(9–10):433–9. http://dx.doi.org/https://doi.org/10.1016/j.drudis.2008.02.001.PubMedView ArticleGoogle Scholar
- Lewis DFV, Dickins M. Substrate SARs in human P450s. Drug Discov Today. 2002;7(17):918–25. http://dx.doi.org/10.1016/S1359-6446(02)02412-1.PubMedView ArticleGoogle Scholar
- Gottesman MM, Fojo T, Bates SE. Multidrug resistance in cancer: role of ATP-dependent transporters. Nat Rev Cancer. 2002;2(1):48–58.PubMedView ArticleGoogle Scholar
- Li X-Z, Nikaido H. Efflux-mediated drug resistance in bacteria. Drugs. 2004;64(2):159–204. doi:https://doi.org/10.2165/00003495-200464020-00004.PubMedView ArticleGoogle Scholar
- Pfaller MA. Antifungal drug resistance: mechanisms, epidemiology, and consequences for treatment. Am J Med. 2012;125(1, Supplement):S3–13. http://dx.doi.org/https://doi.org/10.1016/j.amjmed.2011.11.001.PubMedView ArticleGoogle Scholar
- Gottesman MM. Mechanisms of cancer drug resistance. Ann Rev Med. 2002;53:615–27.PubMedView ArticleGoogle Scholar
- Walsh C. Molecular mechanisms that confer antibacterial drug resistance. Nature. 2000;406(6797):775–81.PubMedView ArticleGoogle Scholar
- Proudfoot JR. Drugs, leads and drug-likeness: an analysis of some recently launched drugs. Bioorg Med Chem Lett. 2002;12:1647–50.PubMedView ArticleGoogle Scholar
- Hann MM, Oprea TI. Pursuing the leadlikeness concept in pharmaceutical research. Curr Opin Chem Biol. 2004;8:255–63.PubMedView ArticleGoogle Scholar
- Oprea TI, Davis AM, Teague SJ, Leeson PD. Is there a difference between leads and drugs? A historical perspective. J Chem Inform Comput Sci. 2001;41:1308–15.View ArticleGoogle Scholar
- Wunberg T, Hendrix M, Hillisch A, Lobell M, Meier H, Schmeck C, et al. Improving the hit-to-lead process: data-driven assessment of drug-like and lead-like screening hits. Drug Discov Today. 2006;11(3-4):175–80.PubMedView ArticleGoogle Scholar
- Lee SK, Chou H, Ham TS, Lee TS, Keasling JD. Metabolic engineering of microorganisms for biofuels production: from bugs to synthetic biology to fuels. Curr Opin Biotechnol. 2008;19(6):556–63. http://dx.doi.org/https://doi.org/10.1016/j.copbio.2008.10.014.PubMedView ArticleGoogle Scholar
- Carbonell P, Fichera D, Pandit SB, Faulon J-L. Enumerating metabolic pathways for the production of heterologous target chemicals in chassis organisms. BMC Systems Biology. 2012;6(10). doi:https://doi.org/10.1186/752-0509-6-10.
- Deplazes A, Huppenbauer M. Synthetic organisms and living machines. Syst Synth Biol. 2009;3(1–4):55–63. doi:https://doi.org/10.1007/s11693-009-9029-4.PubMed CentralPubMedView ArticleGoogle Scholar
- Aubel D, Fussenegger M. Mammalian synthetic biology - from tools to therapies. BioEssays. 2009;32:332–45.View ArticleGoogle Scholar
- Dietz S, Panke S. Microbial systems engineering: first successes and the way ahead. BioEssays. 2009;32:356–62.View ArticleGoogle Scholar
- NCI. NCI Yeast Anticancer Drug Screen. 2013. https://wiki.nci.nih.gov/display/NCIDTPdata/Yeast+Anticancer+Drug+Screen. Accessed 6/2/2013.
- Weininger D. SMILES, a chemical language and information system. 1. Introduction to methodology and coding rules. J Chem Inform Comput Sci. 1988;28:31–6.View ArticleGoogle Scholar
- QB64. QB64 compiler. 2015. http://www.qb64.net/). Accessed 4/1/2015.
- Trott O, Olson AJ. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization and multithreading. J Comput Chem. 2010;31:455–61.PubMed CentralPubMedGoogle Scholar
- Bains W. What do we think life is? A simple illustration and its consequences. International Journal of Astrobiology. 2013;in press.Google Scholar
- USGS. USGS acute toxicity database. 2013. http://www.cerc.usgs.gov/data/acute/acute.html. Accessed 12/9/2013.
- EPA. EPAFHM: EPA Fathead Minnow Acute Toxicity. 2013. http://www.epa.gov/ncct/dsstox/sdf_epafhm.html. Accessed 1/12/2013.
- Budavari S, O'Neil M, Smith A, Heckelman P, Obenchain J, editors. The Merck index. Boca Raton, FL, USA: CRC Press; 1996.Google Scholar
- Toxnet. Toxnet database. 2006. http://toxnet.nlm.nih.gov/. Accessed 3/1/2006.
- EPA. CPDBAS: Carcinogenic Potency Database. 2009. http://www.epa.gov/ncct/dsstox/sdf_cpdbas.html. 11/11/2009.
- NTP. Drosophila testing program. 2013. http://ntp.niehs.nih.gov/testing/types/genetic/invivo/dl/index.html. Accessed 30/1/2013.
- Regional IPM Centers. OPP Pesticide Ecotoxicity Database. 2013. http://www.ipmcenters.org/Ecotox/DataAccess.cfm Accessed 22/11/2013.
- Larson JH, Frost PC, Lamberti GA. Variable toxicity of ionic liquid-forming chemicals to Lemna minor and the influence of dissolved organic matter. Environ Toxicol Chem. 2008;27(3):676–81.PubMedView ArticleGoogle Scholar
- Blackman GE, Parke MH, Garton G. The physiological activity of substituted phenols. I. Relationships between chemical structure and physiological activity. Arch Biochem Biophys. 1955;54(1):45–54. http://dx.doi.org/10.1016/0003-9861(55)90007-2.PubMedView ArticleGoogle Scholar
- Wang W. Literature review on duckweed toxicity testing. Environ Res. 1990;52(1):7–22. http://dx.doi.org/10.1016/S0013-9351(05)80147-1.PubMedView ArticleGoogle Scholar
- Cowgill UM, Milazzo DP, Landenberger BD. The sensitivity of Lemna gibba G-3 and four clones of lemna minor to eight common chemicals using a 7-day test. Res J Water Pollut Fed. 1991;63(7):991–8.Google Scholar
- Tong Z, Hongjun J. Use of Duckweed (Lemna minor L.) growth inhibition test to evaluate the toxicity of acrylonitrile, sulphocyanic sodium and acetonitrile in China. Environ Pollut. 1997;98(2):143–7. http://dx.doi.org/10.1016/S0269-7491(97)00156-5.View ArticleGoogle Scholar
- Hanson ML, Solomon KR. Haloacetic acids in the aquatic environment. Part I: macrophyte toxicity. Environ Pollut. 2004;130(3):371–83. http://dx.doi.org/10.1016/j.envpol.2003.12.016.PubMedView ArticleGoogle Scholar
- Brain RA, Johnson DJ, Richards SM, Sanderson H, Sibley PK, Solomon KR. Effects of 25 pharmaceutical compounds to Lemna gibba using a seven-day static-renewal test. Environ Toxicol Chem. 2004;23(2):371–82. doi:https://doi.org/10.1897/02-576.PubMedView ArticleGoogle Scholar
- Sharma HA, Barber JT, Ensley HE, Polito MA. A comparison of the toxicity and metabolism of phenol and chlorinated phenols by Lemna gibba, with special reference to 2,4,5-trichlorophenol. Environ Toxicol Chem. 1997;16(2):346–50. doi:https://doi.org/10.1002/etc.5620160233.View ArticleGoogle Scholar
- Pillard DA, DuFresne DL. Toxicity of formulated glycol deicers and ethylene and propylene glycol to lactuca sativa, Lolium perenne, Selenastrum capricornutum, and Lemna minor. Arch Environ Contam Toxicol. 1999;37(1):29–35. doi:https://doi.org/10.1007/s002449900486.PubMedView ArticleGoogle Scholar
- Ramirez Toro GI, Leather GR, Einhellig FA. Effects of three phenolic compounds onLemna gibba G3. J Chem Ecol. 1988;14(3):845–53. doi:https://doi.org/10.1007/bf01018778.PubMedView ArticleGoogle Scholar
- Kirby MF, Sheahan DA. Effects of atrazine, isoproturon, and mecoprop on the macrophyte Lemna minor and the alga Scenedesmus subspicatus. Bull Environ Contam Toxicol. 1994;53(1):120–6. doi:https://doi.org/10.1007/bf00205148.PubMedView ArticleGoogle Scholar
- van de Plassche EJ, de Bruijn JHM, Stephenson RR, Marshall SJ, Feijtel TCJ, Belanger SE. Predicted no-effect concentrations and risk characterization of four surfactants: Linear alkyl benzene sulfonate, alcohol ethoxylates, alcohol ethoxylated sulfates, and soap. Environ Toxicol Chem. 1999;18(11):2653–63. doi:https://doi.org/10.1002/etc.5620181135.View ArticleGoogle Scholar
- Boudreau TM, Sibley PK, Mabury SA, Muir DGC, Solomon KR. Laboratory Evaluation of the Toxicity of Perfluorooctane Sulfonate (PFOS) on Selenastrum capricornutum, Chlorella vulgaris, Lemna gibba, Daphnia magna, and Daphnia pulicaria. Arch Environ Contam Toxicol. 2003;44(3):0307–13. doi:https://doi.org/10.1007/s00244-002-2102-6.View ArticleGoogle Scholar
- Qi P, Wang Y, Mu J, Wang J. Aquatic predicted no-effect-concentration derivation for perfluorooctane sulfonic acid. Environ Toxicol Chem. 2011;30(4):836–42. doi:https://doi.org/10.1002/etc.460.PubMedView ArticleGoogle Scholar
- McConkey BJ, Duxbury CL, Dixon DG, Greenberg BM. Toxicity of a pah photooxidation product to the bacteria Photobacterium phosphoreum and the duckweed Lemna gibba: Effects of phenanthrene and its primary photoproduct, phenanthrenequinone. Environ Toxicol Chem. 1997;16(5):892–9. doi:https://doi.org/10.1002/etc.5620160508.View ArticleGoogle Scholar
- Xu Y, Lay JP, Korte F. Fate and effects of xanthates in laboratory freshwater systems. Bull Environ Contam Toxicol. 1988;41(4–6):683–9. doi:https://doi.org/10.1007/bf02021019.PubMedView ArticleGoogle Scholar
- Brain RA, Johnson DJ, Richards SM, Hanson ML, Sanderson H, Lam MW, et al. Microcosm evaluation of the effects of an eight pharmaceutical mixture to the aquatic macrophytes Lemna gibba and Myriophyllum sibiricum. Aquat Toxicol. 2004;70(1):23–40. http://dx.doi.org/10.1016/j.aquatox.2004.06.011.PubMedView ArticleGoogle Scholar
- Berends AG, Boutonnet JC, Rooij CGD, Thompson RS. Toxicity of trifluoroacetate to aquatic organisms. Environ Toxicol Chem. 1999;18(5):1053–9. doi:https://doi.org/10.1002/etc.5620180533.View ArticleGoogle Scholar
- Caux PY, Weinberger P, Carlisle DB. A physiological study of the effects of triton surfactants on Lemna minor L. Environ Toxicol Chem. 1988;7(8):671–6. doi:https://doi.org/10.1002/etc.5620070808.View ArticleGoogle Scholar
- Schultz TW. Structure−toxicity relationships for benzenes evaluated with tetrahymena pyriformis. Chem Res Toxicol. 1999;12(12):1262–7. doi:https://doi.org/10.1021/tx9900730.PubMedView ArticleGoogle Scholar
- Schultz TW, Netzeva TI, Roberts DW, Cronin MTD. Structure−toxicity relationships for the effects to tetrahymena pyriformis of aliphatic, carbonyl-containing, α, β-unsaturated chemicals. Chem Res Toxicol. 2005;18(2):330–41. doi:https://doi.org/10.1021/tx049833j.PubMedView ArticleGoogle Scholar
- Schultz TW, Cronin MTD, Netzeva TI, Aptula AO. Structure−toxicity relationships for aliphatic chemicals evaluated with tetrahymena pyriformis. Chem Res Toxicol. 2002;15(12):1602–9. doi:https://doi.org/10.1021/tx025589p.PubMedView ArticleGoogle Scholar
- Akers KS, Sinks GD, Schultz TW. Structure–toxicity relationships for selected halogenated aliphatic chemicals. Environ Toxicol Pharmacol. 1999;7(1):33–9. http://dx.doi.org/10.1016/S1382-6689(98)00048-9.PubMedView ArticleGoogle Scholar
- Cronin MTD, Netzeva TI, Dearden JC, Edwards R, Worgan ADP. Assessment and modeling of the toxicity of organic chemicals to chlorella vulgaris: development of a novel database. Chem Res Tox. 2004;17:545–54.View ArticleGoogle Scholar
- Yan X-F, Xiao H-M, Gong X-D, Ju X-H. Quantitative structure–activity relationships of nitroaromatics toxicity to the algae (Scenedesmus obliguus). Chemosphere. 2005;59(4):467–71. http://dx.doi.org/10.1016/j.chemosphere.2005.01.085.PubMedView ArticleGoogle Scholar
- Saçan MT, Özkul M, Erdem SS. QSPR analysis of the toxicity of aromatic compounds to the algae (Scenedesmus obliquus). Chemosphere. 2007;68(4):695–702. http://dx.doi.org/10.1016/j.chemosphere.2007.01.043.PubMedView ArticleGoogle Scholar
- Wang C, Lu G, Tang Z, Guo X. Quantitative structure-activity relationships for joint toxicity of substituted phenols and anilines to Scenedesmus obliquus. J Environ Sci. 2008;20(1):115–9. http://dx.doi.org/10.1016/S1001-0742(08)60018-2.View ArticleGoogle Scholar
- Lu G-H, Yuan X, Zhao Y-H. QSAR study on the toxicity of substituted benzenes to the algae (Scenedesmus obliquus). Chemosphere. 2001;44(3):437–40. http://dx.doi.org/10.1016/S0045-6535(00)00214-9.PubMedView ArticleGoogle Scholar
- Ma J, Zheng R, Xu L, Wang S. Differential sensitivity of two green algae, scenedesmus obliqnus and chlorella pyrenoidosa, to 12 pesticides. Ecotoxicol Environ Saf. 2002;52(1):57–61. http://dx.doi.org/10.1006/eesa.2002.2146.PubMedView ArticleGoogle Scholar
- Tadros MG, Philips J, Patel H, Pandiripally V. Differential response of green algal species to solvents. Bull Environ Contam Toxicol. 1994;52(3):333–7. doi:https://doi.org/10.1007/bf00197817.PubMedView ArticleGoogle Scholar
- Zhang W, Zhang M, Lin K, Sun W, Xiong B, Guo M, et al. Eco-toxicological effect of Carbamazepine on Scenedesmus obliquus and Chlorella pyrenoidosa. Environ Toxicol Pharmacol. 2012;33(2):344–52. http://dx.doi.org/10.1016/j.etap.2011.12.024.PubMedView ArticleGoogle Scholar
- Ma J, Lu N, Qin W, Xu R, Wang Y, Chen X. Differential responses of eight cyanobacterial and green algal species, to carbamate insecticides. Ecotoxicol Environ Saf. 2006;63(2):268–74. http://dx.doi.org/10.1016/j.ecoenv.2004.12.002.PubMedView ArticleGoogle Scholar
- Ma J. Differential sensitivity of three cyanobacterial and five green algal species to organotins and pyrethroids pesticides. Sci Total Environ. 2005;341(1–3):109–17. http://dx.doi.org/10.1016/j.scitotenv.2004.09.028.PubMedView ArticleGoogle Scholar
- Ma J, Wang P, Chen J, Sun Y, Che J. Differential response of green algal species Pseudokirchneriella subcapitata, Scenedesmus quadricauda, Scenedesmus obliquus, Chlorella vulgaris and Chlorella pyrenoidosa to six pesticides. Pol J Environ Stud. 2007;16(6):847–51.Google Scholar
- Kulacki KJ, Lamberti GA. Toxicity of imidazolium ionic liquids to freshwater algae. Green Chem. 2008;10:104–10.View ArticleGoogle Scholar
- Geoffroy L, Teisseire H, Couderchet M, Vernet G. Effect of oxyfluorfen and diuron alone and in mixture on antioxidative enzymes of Scenedesmus obliquus. Pestic Biochem Physiol. 2002;72(3):178–85. http://dx.doi.org/10.1016/S0048-3575(02)00009-3.View ArticleGoogle Scholar
- Li X, Ping X, Xiumei S, Zhenbin W, Liqiang X. Toxicity of cypermethrin on growth, pigments, and superoxide dismutase of Scenedesmus obliquus. Ecotoxicol Environ Saf. 2005;60(2):188–92. http://dx.doi.org/10.1016/j.ecoenv.2004.01.012.PubMedView ArticleGoogle Scholar
- EPA. NCTRER: FDA National Center for Toxicological Research Estrogen Receptor Binding Database File. 2013. http://www.epa.gov/ncct/dsstox/sdf_nctrer.html. Accessed 11/2/2013.
- Garg R, Kurup A, Mekapati SB, Hansch C. Cyclooxygenase (COX) inhibitors: a comparative QSAR study. Chem Rev. 2003;103:703–31.PubMedView ArticleGoogle Scholar
- Anthes JC, Gilchrest H, Richard C, Eckel S, Hesk D, West REJ, et al. Biochemical characterization of desloratadine, a potent antagonist of the human histamine H1 receptor. Eur J Pharmacol. 2002;449:229–37.PubMedView ArticleGoogle Scholar
- Kubo N, Shirakawa O, Kuno T, Tanaka C. Antimuscarinic effects of antihistamines: Quantitative evaluation by receptor-binding assay. Japan J Pharmacol. 1987;43:277–82.View ArticleGoogle Scholar
- Jongejan A, Leurs R. Delineation of receptor-ligand interactions at the human histamine H1 receptor by a combined approach of site-directed mutagenesis and computational techniques or how to bind the H1 receptor. Arch Pharm Chem Life Sci. 2005;338:248–59.View ArticleGoogle Scholar