Detecting and understanding meaningful cancerous mutations based on computational models of mRNA splicing - Article highlights

Sep 21, 2026

Uploaded by Finn

Cancer research has long focused on non-silent mutations, but silent and apparently silent changes can influence gene expression and cancer cell fitness through regulatory processes, especially alternative mRNA splicing. This work introduces Oncosplice, a computational tool that scores mutations using neural network splice-site predictions, a greedy transcript constructor, and evolutionary-conservation-based measures of protein functional divergence. Applied to 12M somatic mutations, it identifies deleterious variants depleted in healthy populations and detects clinically validated pathogenic variants with high predictive value, also improving patient survival estimation using novel cancer-associated genes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Background: silent mutations and splicing

Oncosplice: computational pipeline and scoring strategy

Splice-site prediction and transcript construction

Functional divergence and pathogenicity scoring

Applications and validation on somatic and clinical datasets

Results: depleted deleterious variants and clinical performance

Patient survival estimation and gene prioritization

npj | systems biology and applications Article | Published in partnership with the Systems Biology Institute | | | | --- | --- | --- | | [https://doi.org/10.1038/s41540-024-00351-7](https://doi.org/10.1038/s41540-024-00351-7) | | | | Detecting and understanding meaningful cancerous mutations based on computational models of mRNA splicing
Check for updates | | | | Nicolas Lynn & Tamir Tuller | | | | Cancer research has long relied on non-silent mutations. Yet, it has become overwhelmingly clear that silent mutations can affect gene expression and cancer cell ﬁtness. One fundamental mechanism that apparently silent mutations can severely disrupt is alternative splicing. Here we introduce Oncosplice , a tool that scores mutations based on models of proteomes generated using aberrant splicing predictions. Oncosplice leverages a highly accurate neural network that predicts splice sites within arbitrary mRNA sequences, a greedy transcript constructor that considers alternate arrangements of splicing blueprints, and an algorithm that grades the functional divergence between proteins based on evolutionary conservation. By applying this tool to 12M somatic mutations we identify 8K deleterious variants that are signiﬁcantly depleted within the healthy population; we demonstrate the tool’s ability to identify clinically validated pathogenic variants with a positive predictive value of 94%; we show strong enrichment of predicted deleterious mutations across pan-cancer drivers. We also achieve improved patient survival estimation using a proposed set of novel cancer-involved genes. Ultimately, this pipeline enables accelerated insight-gathering of sequence-speciﬁc consequences for a class of understudied mutations and provides an efﬁcient way of ﬁltering through massive variant datasets – functionalities with immediate experimental and clinical applications. | | | | Advancementsin sequencing technologyhave made extensive collections of mutations and genomic information available1–4. These datasets include millions ofnovel mutations that cannot all be experimentally studied due to time and cost constraints. Thus, most investigations that characterize functional variants focus on non-silent, non-synonymous mutations5,6. This heuristic effectively narrows the search space yet neglects thousands of apparently silent mutations with measurable and potentially more severe consequences. Instead of directly altering coding nucleotides, silent and apparently silent mutations act on regulatory gene expression processes6–10; they can exist within introns and untranslated regions, or within coding sequences (CDS)3–11, and they hold signiﬁcant predictive power in cancer classiﬁcation and prognosis11. Among the regulatory mechanisms that they can hijack is splicing9, 12–25.
mRNA splicing is a co-transcriptional modiﬁcation step that transforms one pre-mRNA sequence into multiple transcripts through alternative splicing (AS). The most crucial cis-elements needed for proper splicing are the intron’s5′ (donor–GU motif)and3′ (acceptor–AG motif) junctions. There are also hundreds if not thousands of sequence determinants far within and beyond the intron that are moredifﬁcult to characterize | and play roles of varying importance in deciding which GU/AG dinucleotides serve as functioning splice sites26. Ultimately, this gives cancerous apparently silent mutations countless targets through which they can disrupt healthy gene expression.
Numerous examples illustrate the impact of aberrant splicing in cancer15,27–29. One estimate claims that between 15% and 50% of human disease mutations can result in splicing dysregulation24. It was found that 68% of tumor samples contained at least one aberrant splicing-derived neoepitope while only 30% contained neoepitopes derived from somatic single-nucleotide variants30. Exons 4, 6, and 9 of TP53 contain functional hotspots for intron retention-caused inactivation by SNVs, and mutations causing such e

FAQs

Why do silent mutations matter in cancer research?

Silent and apparently silent mutations can affect gene expression and cancer cell fitness by disrupting regulatory processes, particularly alternative splicing, rather than only altering protein-coding nucleotides.

What is Oncosplice and what does it score?

Oncosplice predicts aberrant splicing events, constructs variant transcriptome and proteomes, and assigns pathogenicity scores based on estimated functional divergence between reference and aberrant proteins.

How was Oncosplice evaluated and what were the key outcomes?

Applied to 12M somatic mutations, it identifies deleterious variants depleted in healthy populations, detects clinically validated pathogenic variants with a reported positive predictive value of 94%, and shows enrichment of deleterious predictions across pan-cancer drivers.

Sep 21, 2026

Uploaded by Finn