Nakatsui, M. | Search bioRxiv

bioRxiv2024

Genome-scale prediction of gene ontology from mass fingerprints reveals new metabolic gene functions

Mass-based fingerprinting can characterize unknown strains, however expansion of these methods to predict specific gene functions is lacking. Therefore, rapid mass fingerprinting was developed to functionally profile a comprehensive yeast knockout library. Matrix assisted laser desorption ionization (MALDI)-time of flight (TOF) mass fingerprints of 3,238 Saccharomyces cerevisiae knockouts were digitized for correlation with gene ontology (GO) annotations. Random forests and support vector machine (SVM) algorithms precisely assigned GO accessions with AUC scores all above 0.83. SVM was the best predictor with average true positive and true negative rates of 0.975 and 0.991, respectively. The SVM model suggested new functions for 28 uncharacterized yeast genes. Metabolomics analysis of two knockouts (YDR215C and YLR122C) of uncharacterized genes predicted to be involved in methylation-related metabolism, showed altered intracellular contents of methionine-related metabolites. Increased S-adenosylmethionine in YDR215C highlights potential for enhancement of methylation pathways. These results demonstrate that MALDI-TOF fingerprints can be rapidly digitized, resulting in datasets that enable prediction of microbial genotypes and even the function of specific genes. This fingerprinting method can inform optimal bioproduction chassis selection.

systems biology↗

bioRxiv2021

Network-based pathogenicity prediction for variants of uncertain significance

While the number of genome sequences continues to increase, the functions of many detected gene variants remain to be identified. These variants of uncertain significance constitute a major barrier to precision medicine 1-3. Although many computational methods have been developed to predict the function of these variants, they all rely on individual gene features and do not consider complex molecular relationships. Here we develop PathoGN, a molecular network-based approach for predicting variant pathogenicity. PathoGN significantly outperforms existing methods using benchmark datasets. Moreover, PathoGN successfully predicts the pathogenicity of 3,994 variants of uncertain significance in the real-world database ClinVar and designates potential pathogenicity. This is the first computational method for the clinical interpretation of variants using biomolecular networks, and we anticipate our method to be broadly useful for the clinical interpretation of variants and for assigning biological function to unknown variants at the genomic scale.

genomics↗