Fig. 1
From: Inferring phenotypes from substance use via collaborative matrix completion
Inferring phenotypes for diagnoses of substance use disorders (e.g., opioid and cocaine illustrated here) via matrix completion. Phenotypes for a set of patients are organized into a matrix F with rows for m patients and columns for n diagnostic symptoms. Related features that describe patients and symptoms are available, such as, genotypes of individuals in the matrix X and pair-wise similarities between diagnostic symptoms in the matrix Y. A bi-linear model: F=XTGY+N is used where G and N are model parameters to be learned, and quantify the impact of genotype-symptom interactions and the residual from any other effect (mainly random environmental effect), respectively