Bias in random forest variable importance measures: Illustrations,sources and a solution

Background  

Variable importance measures for random forests have been receiving increased attention as a means of variable selection in many classification tasks in bioinformatics and related scientific fields, for instance to select a subset of genetic markers relevant for the prediction of a certain disease. We show that random forest variable importance measures are a sensible means for variable selection in many applications, but are not reliable in situations where potential predictor variables vary in their scale of measurement or their number of categories. This is particularly important in genomics and computational biology, where predictors often include variables of different types, for example when predictors include both sequence data and continuous variables such as folding energy, or when amino acid sequence data show different numbers of categories.     

The behaviour of random forest permutation-based variable importance measures under predictor correlation

Background  

Random forests (RF) have been increasingly used in applications such as genome-wide association and microarray studies where predictor correlation is frequently observed. Recent works on permutation-based variable importance measures (VIMs) used in RF have come to apparently contradictory conclusions. We present an extended simulation study to synthesize results.     

Erratum to: Informative metabolites identification by variable importance analysis based on random variable combination

Metabolomics - The quality of rice in terms not only of its nutritional value but also in terms of its aroma and flavour is becoming increasingly important in modern rice breeding where global...     

Letter to the editor: Non steroid anti-inflammatory drugs and pregnancy

A review of relevant clinical research literature leads to the conclusion that any known nonsteroid anti-inflammatory drug (NSAID) should be contraindicated for use during pregnancy. This conclusion is reached as a result of the potentially profound effects of NSAIDs noted on platelet functions and their inhibition of specific tissue prostaglandin(PG) synthetases. Study results have shown that administration of Naproxen to midtrimester abortion patients prolonged the instillation-abortion interval from 33.4 hours +or- 2.9 hours (in a placebo treated group) to 64.7 hours +or- 6.4 hours. Treatment with aspirin prolonged induction-abortion time by about 9 hours and treatment with indomethacin by 30 hours. 2 recent clinical papers have shown that use of acetylsalicylic acid, a relatively weak inhibitor of PG synthetase when compared to Naproxen, by pregnant women at or near term increased their length of gestation, their mean duration of labor, the frequency of postmaturity among them, and the incidence of pre- and postpartum hemorrhage. A small dose of aspirin given to mothers at or near term results in platelet dysfunction in both the mother and the newborn infant. NSAID most frequently disrupts normal hemostatic homeostasis mechanisms. In addition, studies have shown that use of NSAIDs increases the incidence of chromosomal abnormalities in spontaneous abortuses. Such drugs should not be used to prevent threatened abortions.     

Letter to the editor: Influence of proline hydroxylation upon the thermal stability of collagen fragment alpha1CB2

A polarimetric study of a 36-residue peptide, α1CB2, obtained by CNBr cleavage of rat-tail tendon collagen, has been made. The equilibrium and kinetic data for this peptide are compared with those found for rat-skin collagen peptide α1CB2. The marked difference in thermal stability is interpreted in terms of the differing degree of hydroxylation of proline residues in otherwise identical peptides.