Dataset Information

Machine learning for discovery: deciphering RNA splicing logic

ABSTRACT: Machine learning methods, particularly neural networks trained on large datasets, are transforming how scientists approach scientific discovery and experimental design. However, current state-of-the-art neural networks are limited by their uninterpretability: despite providing accurate predictions, they cannot describe how they arrived at their predictions. Here, using an ``interpretable-by-design'' approach, we present a neural network model that provides insights into RNA splicing, a fundamental process in the transfer of genomic information into functional biochemical products. Although we designed our model to emphasize interpretability, its predictive accuracy is on par with state-of-the-art models. To demonstrate the model's interpretability, we introduce a visualization that, for any given exon, allows us to trace and quantify the entire decision process from input sequence to output splicing prediction. Importantly, the model revealed novel components of the splicing logic, which we experimentally validated. This study highlights how interpretable machine learning can advance scientific discovery.

ORGANISM(S): Homo sapiens

PROVIDER: GSE200096 | GEO | 2022/10/01

REPOSITORIES: GEO

ACCESS DATA

Similar Datasets

Project description:Pre-mRNA splicing is a key control point in human gene expression. Disturbances in splicing due to mutation or aberrant splicing regulatory networks lead to dysregulated protein expression and contribute to a substantial fraction of human disease. Several classes of active and selective splicing modulator compounds (SMCs) have been recently identified and establish that pre-mRNA splicing represents a viable target for therapy. This also raises the intriguing possibility that SMCs may have broad capabilities to ameliorate aberrant splicing across multiple human disorders. We describe herein the identification of BPN-15477, a novel SMC that restores correct splicing of exon 20 in the Elongator complex protein 1 (ELP1) gene carrying the major IVS20+6T>C mutation responsible for familial dysautonomia. Given that BPN-15477 corrects splicing and increases full-length ELP1 protein in vivo, we developed a machine learning approach to evaluate the therapeutic potential of BPN-15477 to correct splicing in other human genetic diseases. Using transcriptome sequencing from compound-treated fibroblast cells, we identified treatment responsive sequence signatures, the majority of which center at the 5' splice site of exons whose inclusion or exclusion is modulated by SMC treatment. We then leveraged this model to identify 155 human disease genes that harbor ClinVar mutations predicted to alter pre-mRNA splicing as potential targets for BPN-15477 treatment. Using in vitro splicing assays, we validated representative predictions by demonstrating successful correction of splicing defects caused by mutations in the genes responsible for cystic fibrosis (CFTR), cholesterol ester storage disease (LIPA), Lynch syndrome (MLH1) and familial frontotemporal dementia (MAPT). Importantly, we also validated these predictions in two disease relevant cellular models for LIPA and CFTR, confirming that treatment increases functional protein and confirming the clinical potential for our model predictions. Our study shows that deep learning techniques can identify a complex set of sequence signatures and predict response to pharmacological modulation, strongly supporting the use of in silico approaches to expand the therapeutic potential of drugs that modulate splicing.

Dataset Information

Machine learning for discovery: deciphering RNA splicing logic

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure