Splicedisco: a machine learning tool for the identification of alternative splicing events directly regulated by pathological splicing factors in cancer

The National Cancer Institute awarded Envisagenics Inc. $889,653 on August 1, 2026, under the Cancer Detection and Diagnosis Research program (CFDA 93.394) to develop SPLICEDISCO, a machine learning algorithm that identifies alternative splicing events directly regulated by pathological splicing factors in cancer.

SPLICEDISCO bridges the gap between dysregulated splicing factors and enriched alternative splicing target candidates to support therapeutic development in breast cancer, non-small cell lung cancer, colorectal cancer, and acute myeloid leukemia. The algorithm establishes direct regulatory associations between pathological splicing factors and alternative splicing targets, improving understanding of disease mechanisms and identifying targets most likely to be effective for immuno-oncology therapeutics. The work builds on Envisagenics' existing commercialized software tools, SPLICECORE, which enables alternative splicing target candidate identification, and SPLICESLICE, which stratifies patients based on splicing factor expression.

Performance takes place in Astoria, New York, with a period of performance from August 1, 2026, through July 31, 2028. The assistance type is a Project Grant, classified as a Small Business Innovation Research (SBIR) Phase II award supporting commercialization of innovations derived from federal research and development.

Abstract

Many cancers are caused or aggravated by RNA splicing errors, which have been shown to have a major impact on cancer progression. Interestingly, while tumor cells often thrive on splicing errors, these can also become their weakness. This is because, tumor-specific alternative splicing (as) events produce unique proteins that can serve as potential targets for immuno-oncology (IO) therapeutics. This project addresses the pressing need to develop effective IO therapeutics for cancers driven by RNA splicing errors, including breast cancer (BC), non- small cell lung cancer (nsclc), colorectal cancer (CRC), and acute myeloid leukemia (AML). Dysregulation of splicing factors (SFS) in cancer leads to the occurrence of tumor-specific as events. While RNA sequencing is a powerful tool for the detection of tumor-specific as events, identifying events directly linked to pathological cancer processes is important, because as events indirectly influenced by cellular changes or with no relevance to cancer processes are unlikely to be good therapeutic targets. This project proposes the development of the splicedisco algorithm to establish direct regulatory associations between pathological SFS and as target candidates, which will facilitate therapeutic development by improving our understanding of disease mechanisms and identifying targets most likely to be effective. Building upon the success of our commercialized software which enables as target candidate identification (splicecore) and patient stratification based on SFS (spliceslice), splicedisco aims to bridge the GAP between pathological SFS and the enrichment of as target candidates, contributing to the commercialization and development of as-targeted drugs. Our existing computational tools have supported the identification of as target candidates 3 times more prevalent in specific patient subgroups with SF abnormalities than the overall population. The development of splicedisco will enable a deep understanding of SF-as regulatory interactions. Splicedisco will be developed and optimized using machine learning models. Splicedisco's performance will be 1) proven using existing computational tools and data and 2) experimentally validated in cancer patient tissues along with cell lines and organoids. The goal of this direct-to-phase II application is to develop and validate splicedisco, a machine learning software designed to identify as target candidates that are directly influenced by pathological SFS. We have already identified numerous as target candidates with splicecore, which have been experimentally validated using our established laboratory validation pipeline. To support our efforts, splicedisco will provide data to understand disease mechanisms and improve our ability to prioritize effective targets, which will streamline the development of our as-targeted drugs for cancer patients. In this proposal, we will 1) develop splicedisco, 2) implement splicedisco across several cancers with abundant tumor-specific as, and 3) experimentally validate splicedisco-identified SF-as regulatory interactions.