Protein domains provide a new layer of information for classifying human variations in rare diseases

Abstract

Introduction: Using the ACMG-AMP guidelines for the interpretation of sequence variants, it remains difficult to meet the criterion associated with the protein domain, PM1, which is assigned in only about 10% of cases, whereas the criteria related to variant frequency, PM2/BA1/BS1, is reported in 50% of cases. To improve the classification of human missense variants using protein domains information, we developed the DOLPHIN system (https://dolphin.mmg-gbit.eu). Methods: We used Pfam alignments of eukaryotes to define DOLPHIN scores to identify protein domain residues and variants that have a significant impact. In parallel, we enriched gnomAD variants frequencies for each domains' residue. These were validated using ClinVar data. Results: We applied this method to all potential human transcripts' variants, resulting in 30.0% being assigned a PM1 label, whereas 33.2% were eligible for a new benign support criterion, BP8. We also showed that DOLPHIN provides an extrapolated frequency for 31.8% of the variants, compared to the original frequency available in gnomAD for 7.6% of them. Discussion: Overall, DOLPHIN allows a simplified use of the PM1 criterion, an expanded application of the PM2/BS1 criteria and the creation of a new BP8 criterion. DOLPHIN could facilitate the classification of amino acid substitutions in protein domains that cover nearly 40% of proteins and represent the sites of most pathogenic variants.

Keywords: ACMG guidelines; BP8; BS1; PM1; PM2; protein domain; variant classification.

FIGURE 1

Distribution of the 9,121 ClinVar variations with a quality of at least two stars located in protein domains. X-axis = DOLPHIN “wt” scores; Y-axis = DOLPHIN “∆” scores. Red triangles = ClinVar Class 4 & 5 variations (n = 4,382). Green dots = ClinVar Class 1 & 2 variations (n = 4,739). Horizontal and vertical bands on X and Y-axis represent the corresponding “wt” and “∆” values of each variation according to their type: red for pathogenic variations (class 4 & 5) and green for neutral variations (class 1 & 2).

FIGURE 2

Variations localized in protein domains with a DOLPHIN frequency greater then 5% and reported in gnomAD. X-axis (logarithmic scale) = gnomAD frequency; Y-axis = DOLPHIN Frequency. Note that only a minority of variations (2.4%) have the same frequency in both datasets.