Table 4 Strengths and weaknesses of the approaches

Association

• Fast in implementation

• No need to pre-determine parameters

•Results can be easily validated by hand calculation

• Unsatisfactory in predictive power

•Does not consider the structures all relevant protein-module associations as a whole

•Do not have control for possible over-prediction of associations between frequently occurring domain-disease pairs

MLE

• Good in predictive power

•Fast in implementation

•Take into account the structures all relevant protein-module associations as a whole

• Need to pre-determine parameters

•Do not have control for possible over-prediction of associations between frequently occurring domain-disease pairs

DPEA

• Satisfactory in predictive power

•No need to pre-determine parameters

•Have control for possible over-prediction of associations between frequently occurring domain-disease pairs

• Slow in implementation when the number of candidate domain-disease associations is large

Bayesian

• Excellent in predictive power

•Take into account the structures all relevant protein-module associations as a whole

• Slow in implementation when the number of candidate domain-disease associations is large

•Do not have control for possible over-prediction of associations between frequently occurring domain-disease pairs

•Failure when the log-concave conditions of parameters are not satisfied

PE

• Satisfactory in predictive power

•Only one pre-determined parameter

•Have control for possible over-prediction of associations between frequently occurring domain-disease pairs

• Slow in implementation when the number of candidate domain-disease associations is large