Machine learning-based identification and characterization of 15 novel pathogenic SUOX missense mutations

Alexander Tobias Kaczmarek; Nike Bahlmann; Besarta Thaqi; Patrick May; Guenter Schwarz

doi:10.1016/j.ymgme.2021.07.011

Machine learning-based identification and characterization of 15 novel pathogenic SUOX missense mutations

Mol Genet Metab. 2021 Sep-Oct;134(1-2):188-194. doi: 10.1016/j.ymgme.2021.07.011. Epub 2021 Aug 8.

Authors

Alexander Tobias Kaczmarek¹, Nike Bahlmann², Besarta Thaqi², Patrick May³, Guenter Schwarz⁴

Affiliations

¹ Institute of Biochemistry, Department of Chemistry, University of Cologne, Germany; Center for Molecular Medicine Cologne (CMMC), University of Cologne, Germany.
² Institute of Biochemistry, Department of Chemistry, University of Cologne, Germany.
³ Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg, Belvaux, Luxembourg.
⁴ Institute of Biochemistry, Department of Chemistry, University of Cologne, Germany; Center for Molecular Medicine Cologne (CMMC), University of Cologne, Germany. Electronic address: gschwarz@uni-koeln.de.

PMID: 34420858
DOI: 10.1016/j.ymgme.2021.07.011

Abstract

Isolated sulfite oxidase deficiency (ISOD) is a rare hereditary metabolic disease caused by absence of functional sulfite oxidase (SO) due to mutations of the SUOX gene. SO oxidizes toxic sulfite and sulfite accumulation is associated with neurological disorders, progressive brain atrophy and early death. Similarities of these neurological symptoms to abundant diseases like neonatal encephalopathy underlines the raising need to increase the awareness for ISOD. Here we report an interdisciplinary approach utilizing exome/genome data derived from gnomAD database as well as published variants to predict the pathogenic outcome of 303 naturally occurring SO missense variants and combining these with activity determination. We identified 15 novel ISOD-causing SO variants and generated a databank of pathogenic SO missense variants to support future diagnosis of ISOD patients. We found six inactive variants (W101G, H118Y, E197K, R217W, S427W, D512Y, Q518R) and seven (D110H, P119S, G121E, G130R, Y140C, R269H, Q396P, R459Q) with severe reduction in activity. Based on the Hardy-Weinberg-equilibrium and the combination of our results with published SO missense and protein truncating variants, we calculated the first comprehensive incidence rate for ISOD of 1 in 1,377,341 births and provide a pathogenicity score to 303 naturally occurring SO missense variants.

Keywords: Isolated sulfite oxidase deficiency; Machine learning; Molybdenum cofactor; Random forest classification; SUOX; Sulfite oxidase.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Brain / pathology
Databases, Genetic*
Gene Frequency
Genetic Variation
Genome
Humans
Infant, Newborn
Machine Learning*
Mutation, Missense*
Oxidoreductases Acting on Sulfur Group Donors / deficiency*
Oxidoreductases Acting on Sulfur Group Donors / genetics*

Substances

Oxidoreductases Acting on Sulfur Group Donors
SUOX protein, human