Modular modeling improves the predictions of genetic variant effects on splicing

Abstract

Predicting the effects of genetic variants on splicing is highly relevant for human genetics. We describe the framework MMSplice (modular modeling of splicing) with which we built the winning model of the CAGI 2018 exon skipping prediction challenge. The MMSplice modules are neural networks scoring exon, intron, and splice sites, trained on distinct large-scale genomics datasets. These modules are combined to predict effects of variants on exon skipping, alternative donor and acceptor sites, splicing efficiency, and pathogenicity, with matched or higher performance than state-of-the-art. Our models, available in the repository Kipoi, apply to variants including indels directly from VCF files.

Assessment of branch point prediction tools to predict physiological branch points and their alteration by variants

“…This last point highlights the complexity of splicing that does not only depend on the 5'ss, 3'ss and the BP. To illustrate this complexity, a recent study was published [23] demonstrating the MMSplice tool which gathers several features from intronic and exonic pre-mRNA sequences. This tool was assayed on the Vex-seq data [24] To investigate potential spliceogenic variants occurring in the BP area, we gathered a large collection of 120 human variants (62 unpublished), with their corresponding in vitro RNA data.…”

Functional Analysis of the PCCA and PCCB Gene Variants Predicted to Affect Splicing

“…Bioinformatic analysis of probable splicing alterations was performed with freely available tools: HumanSplicingFinder 3.1 (HSF3.1) ( ) (accessed on 12 June 2020) [ 20 ], SpliceAI [ 21 ], MMSplice [ 22 ], SpiP [ 23 ], Ex-Skip ( ) (accessed on 15 April 2021) [ 24 ], Hexplorer ( ) (accessed on 15 April 2021) [ 25 ] and HExoSplice ( ) (accessed on 15 April 2021) [ 26 ].…”

Peer Review #2 of "Genomics enters the deep learning era (v0.1)"

“…Agarwal et al predicted gene activity from 10 kbp DNA sequences surrounding the TSS(Agarwal and Shendure (2020)). The authors could not find motifs used by the network but an analysis of the over-represented k-mers in the promoters of highly active genes (according to the network) reveals the importance of CpG islands in predicting gene activity.Splicing, translation and polyadenylation of RNACheng et al developed a neural network to predict gene splice sites from the RNA sequence(Cheng et al (2019(Cheng et al ( , 2021). Analysis of the effect of variants using this network shows its utility in understanding the genomic causes of autism.…”