Prioritizing and characterizing functionally relevant genes across human tissues

Abstract

Knowledge of genes that are critical to a tissue’s function remains difficult to ascertain and presents a major bottleneck toward a mechanistic understanding of genotype-phenotype links. Here, we present the first machine learning model–FUGUE–combining transcriptional and network features, to predict tissue-relevant genes across 30 human tissues. FUGUE achieves an average cross-validation auROC of 0.86 and auPRC of 0.50 (expected 0.09). In independent datasets, FUGUE accurately distinguishes tissue or cell type-specific genes, significantly outperforming the conventional metric based on tissue-specific expression alone. Comparison of tissue-relevant transcription factors across tissue recapitulate their developmental relationships. Interestingly, the tissue-relevant genes cluster on the genome within topologically associated domains and furthermore, are highly enriched for differentially expressed genes in the corresponding cancer type. We provide the prioritized gene lists in 30 human tissues and an open-source software to prioritize genes in a novel context given multi-sample transcriptomic data.

Predicting molecular mechanisms of hereditary diseases by using their tissue‐selective manifestation

“…Per model, ranks were determined based on gene TRACE scores, such that top scoring gene was ranked first. 85% of the verified disease genes ranked above the median; 55% of the verified disease genes ranked at the top quartile; and 34% of the verified disease genes ranked at the top 10%. Comparison between the rank of the verified disease genes out of the patient's candidate disease genes between gnomAD (Karczewski et al , 2020), expression‐based prioritization, GADO (Deelen et al , 2019), FUGUE (Somepalli et al , 2021), and TRACE. Median rank of TRACE: 39; gnomAD pLoF: 61, missense: 60; expression‐based prioritization: 43; GADO: 45; FUGUE: 60.…”

“…TRACE was implemented as an early integration ML scheme that utilized 4,744 tissue-based gene features, which were derived and combined from heterogeneous omics sources. The large variety of gene features greatly exceeded that of previous methods (Somepalli et al, 2021), enhancing TRACE interpretability. We trained and tested TRACE on 18,927 protein-coding genes, including 1,031 disease genes that underlie tissue-selective Mendelian diseases, which manifest in eight main tissues.…”

“…, expression-based prioritization, GADO(Deelen et al, 2019), FUGUE(Somepalli et al, 2021), and TRACE. Median rank of TRACE: 39; gnomAD pLoF: 61, missense: 60; expression-based prioritization: 43; GADO: 45; FUGUE: 60.…”

Ten-year distant-recurrence risk prediction in breast cancer by CanAssist Breast (CAB) in Dutch sub-cohort of the randomized TEAM trial

“…Therefore, it would be much simpler to use and more economically efficient than pure gene-profiling-based prognostic tools [6,16]. It is unique because this test uses immunohistochemistry combined with an SVML-based algorithm to predict risk scores [7,[21][22][23][24][25]. CAB segregates patients into two groups based on what corresponds to the CAB risk score with a cutoff of 15.5 [26][27][28][29][30][31][32].…”

In silico identification of the rare-coding pathogenic mutations and structural modeling of human NNAT gene associated with anorexia nervosa

“…Exploring the tissue speci c gene expression mechanism(s) provides the elaborated basis to understand how tissues are distinguished by gene expression patterns and implying their signi cant regulatory role. The preferred function and expression of particular gene in one or more several tissues (or cells) types set better understanding of genes functionality or tissue -gene relationship, its etiology and discovery of novel tissue-speci c drug targets that are especially studied to manage the gene-associated diseases [49].…”