Metric Learning on Expression Data for Gene Function Prediction

Abstract

Motivation: Co-expression of two genes across different conditions is indicative of their involvement in the same biological process. However, using RNA-Seq datasets with many experimental conditions from diverse sources introduces batch effects and other artefacts that might obscure the real co-expression signal. Moreover, only a subset of experimental conditions is expected to be relevant for finding genes related to a particular Gene Ontology (GO) term. Therefore, we hypothesize that when the purpose is to find similar functioning genes that the co-expression of genes should not be determined on all samples but only on those samples informative for the GO term of interest. Results: To address both types of effects, we developed MLC (Metric Learning for Co-expression), a fast algorithm that assigns a GO-term-specific weight to each expression sample. The goal is to obtain a weighted co-expression measure that is more suitable than the unweighted Pearson correlation for applying Guilt-By-Association-based function predictions. More specifically, if two genes are annotated with a given GO term, MLC tries to maximize their weighted co-expression, and, in addition, if one of them is not annotated with that term, the weighted co-expression is minimized. Our experiments on publicly available Arabidopsis thaliana RNA-Seq data demonstrate that MLC outperforms standard Pearson correlation in term-centric performance.

DeepGOWeb: fast and accurate protein function prediction on the (Semantic) Web

“…Several computational approaches to predicting protein functions have been developed ( 1 ). These approaches rely on different types of information that can be used to predict protein functions, including the protein sequence ( 2 , 3 ), interaction networks ( 4 ), gene expression ( 5 ), sequence similarity ( 6 ), phenotypes resulting from loss of function mutations ( 7 ) and text mining ( 8 ). The different types of information can often provide complementary information and, consequently, combining multiple types of information can often improve predictive performance ( 1 ).…”

Robust gene coexpression networks using signed distance correlation

“…Representing gene coexpression as networks eases the study and visualisation of the expression data (Weirauch, 2011;Magwene and Kim, 2004). One motivation behind creating these networks is that genes which are coexpressed across multiple samples are likely to have related functions (Hughes et al, 2000;Stuart et al, 2003;van Noort et al, 2003;Makrodimitris et al, 2020), allowing inference of gene function using guilt by association approaches (Wolfe et al, 2005). This procedure is especially useful if the studied organism is poorly annotated.…”

Solid-State NMR and MD Study of the Structure of the Statherin Mutant SNa15 on Mineral Surfaces

“…[11][12][13][14][15][16][17][18] These protein-surface systems include either small extracellular matrix (ECM) proteins or ECM protein domains that are adsorbed onto hydroxyapatite (HAP) crystals. 14,[19][20][21][22][23][24][25][26][27][28][29] These studies suggest that protein secondary structure can change when adsorbed to surfaces. They also hypothesize that the interaction of amino acid side-chains with inorganic surfaces are dependent on the secondary structures of adsorbed proteins.…”

“…They also hypothesize that the interaction of amino acid side-chains with inorganic surfaces are dependent on the secondary structures of adsorbed proteins. 20,22,24,25,27,30 Thus, all components of protein-surface interaction can be determined only when the structure of the adsorbed protein is resolved.…”

“…Notably, Rosetta and RosettaSurface modeling programs have been used for this. 22,32,42 However, these allatom models of the protein include simplifying assumptions that make calculations faster, like freezing the surface, restraining the protein backbone, and not accounting for waters and ions. Thus, they might exclude important interactions that affect protein structures.…”