[genome-announce] Deleteriousness Predictions: ClinPred for hg19 and hg38
Jairo Navarro Gonzalez
We are pleased to announce the release of the ClinPred pathogenicity score track for hg19 and hg38. ClinPred is a machine-learning predictor of pathogenicity for nonsynonymous (missense) single-nucleotide variants, combining existing pathogenicity scores with population allele frequency from gnomAD. It was trained on confidently annotated disease-causing and benign variants from ClinVar. Pre-computed scores are provided for all possible human missense variants in the exome.
Scores range from 0 to 1, with higher values indicating a greater predicted likelihood that a variant is disease-relevant. The authors recommend a score of ≥ 0.5 as evidence of pathogenicity. As with any pathogenicity prediction score, ClinPred is intended as supporting evidence rather than a stand-alone classifier.
There are four subtracks in this collection, one for each possible alternate nucleotide (A, C, G, T). At every exome position covered by ClinPred, three of the four subtracks show a score (one per non-reference base), and the fourth, corresponding to the reference base, is set to 0. Synonymous alternates are also set to 0, since ClinPred only scores missense variants. Positions with no exome coverage are shown as gaps.
ClinPred pathogenicity scores for all possible single-nucleotide substitutions (A, C, G, and T) at a genomic locus (chr7, GRCh38/hg38). The four ClinPred subtracks display predicted pathogenicity scores for each possible alternate allele, where higher scores indicate a greater likelihood that the variant is disease-causing.
Items in this track are colored according to score:
- Red – Likely pathogenic (≥ 0.5)
- Blue – Likely benign (< 0.5)
Note: Zoom in until every base is visible at the top of the display; otherwise, multiple nucleotides will fall under a single pixel, and no score will be shown on the mouseover tooltip.
We would like to thank the ClinPred authors for making the pre-computed scores publicly available. This track was developed by Lou Nassar and Max Haeussler with QA by Eliza Alde, Barali Kitiyakara, and Jairo Navarro.