Fig. 1
From: RExPRT: a machine learning tool to predict pathogenicity of tandem repeat loci
Methodology of RExPRT. RExPRT was trained on 40 known pathogenic TRs and 745 benign TRs that are commonly expanded in the 1000 Genomes Project controls. These TRs were annotated with features, which are used in a supervised statistical learning approach to classify TRs as pathogenic or benign. Seven different models were trained and validated using the LOOCV technique. Two models were selected and fine-tuned to create an optimized ensemble method for ranking repeats. Twenty one pathogenic TRs and 83 rare, benign TRs were used for testing RExPRT’s performance