In Silico Binary Classification of Catalytic Function (Unaffected or Impaired) of Flavin-Containing Monooxygenase 3 (FMO3) Variants and in Vitro Validation Using New Variants Found in a Japanese Genome Resource Database

Abstract

Functionally impaired human flavin-containing monooxygenase 3 (FMO3) variants are associated with metabolic trimethylaminuria. The present study predicted the level of impairment of catalytic function of eight FMO3 variants newly detected in the updated Tohoku Medical Megabank. Catalytic function was modeled using the in vitro intrinsic trimethylamine N-oxygenation activities of 108 FMO3 variants using a logistic linear regression model. FMO3 proteins with single, double, triple, and quadruple amino acid variants underwent binary classification as unaffected (score 0) or impaired (≤70% of wild-type FMO3, score 1) based on their trimethylamine N-oxygenation activities. Using a model with the sum of the coefficients before and after amino acid substitutions, the mean probabilities of 62 FMO3 variants with impaired catalytic activities (0.78) were significantly higher than those of 46 with unaffected catalytic activities (0.30) (p < 0.01). The area under the corresponding receiver operating characteristic curve was 0.91. This in silico evaluation was applied to eight FMO3 variants (p.Met82Thr, p.Gly180Val, p.Ile212Leu, p.Asn275Ile, p.Gly276Glu, p.Ala377Asp, p.Tyr469Cys, and p.Val502Gly) newly found in the Tohoku Medical Megabank and validated in vitro. The present model could prove useful for estimating the activities of new FMO3 variants by applying a logistic linear regression model based on in vitro-generated catalytic data.