Open AccessA survival of the fittest strategy for the selection of genotypes by which drug responders and non-responders can be predicted in small groups
Author(s) -
Daniël Höhle,
Kim van Rooij,
Jos Bloemers,
James G. Pfaus,
Frits Michiels,
Paddy K C Janssen,
Eric Claassen,
Adriaan Tuiten
Publication year - 2021
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0246828
Subject(s) - single nucleotide polymorphism , medicine , placebo , genotype , biology , genetics , gene , pathology , alternative medicine
Phenotype Prediction Scores (PPS) might be powerful tools to predict traits or the efficacy of treatments based on combinations of Single-Nucleotide Polymorphism ( SNPs) in large samples. We developed a novel method to produce PPS models for small samples sizes. The set of SNPs is first filtered on those known to be relevant in biological pathways involved in a clinical condition, and then further filtered repeatedly in a survival strategy to select stabile positive/negative risk alleles. This method is applied on Female Sexual Interest/Arousal Disorder (FSIAD), for which two subtypes has been proposed: 1) a relatively insensitive excitatory system in the brain for sexual cues, and 2) a dysfunctional activation of brain mechanisms for sexual inhibition. A double-blind, randomized, placebo-controlled cross-over experiment was conducted on 129 women with FSIAD. The women received three different on-demand drug-combination treatments during 3 two-week periods: testosterone (0.5 mg) + sildenafil (50 mg), testosterone (0.5 mg) + buspirone (10 mg), or matching placebos. The resulted PPS were independently validated on patient-level and group-level. The AUC scores for T+S of the derivation set was 0.867 (95% CI = 0.796–0.939; p<0.001) and was 0.890 (95% CI = 0.778–1.000; p<0.001) on the validation set. For T+B the AUC of the derivation set was 0.957 (95% CI = 0.921–0.992; p<0.001) and 0.869 (95% CI = 0.746–0.992; p<0.001) for the validation set. Both formulas could reliably predict for each drug who benefit from the on-demand drugs and could therefore be useful in clinical practice.