Aberrant splicing prediction across human tissues

Abstract

Aberrant splicing is a major cause of genetic disorders but its direct detection in transcriptomes is limited to clinically accessible tissues such as skin or body fluids. While DNA-based machine learning models can prioritize rare variants for affecting splicing, their performance in predicting tissue-specific aberrant splicing remains unassessed. Here we generated an aberrant splicing benchmark dataset, spanning over 8.8 million rare variants in 49 human tissues from the Genotype-Tissue Expression (GTEx) dataset. At 20% recall, state-of-the-art DNA-based models achieve maximum 12% precision. By mapping and quantifying tissue-specific splice site usage transcriptome-wide and modeling isoform competition, we increased precision by threefold at the same recall. Integrating RNA-sequencing data of clinically accessible tissues into our model, AbSplice, brought precision to 60%. These results, replicated in two independent cohorts, substantially contribute to noncoding loss-of-function variant identification and to genetic diagnostics design and analytics.

Analysis of 3,760 hematologic malignancies reveals rare transcriptomic aberrations of driver genes

“…We next annotated the genes using the seven features from the software intOGen, which include positional recurrence of mutations in genome sequence (OncodriveCLUSTL), positional recurrence of mutations in protein conformation (HotMAPS), enrichment of mutations in functional domains (smRegions), three alternative measures of selection strength inferred from synonymous and nonsynonymous mutations (CBaSE, MutPanning, and dNdScv), and OncodriveFML, a method identifying excess of somatic mutations across tumors in both coding and non-coding genomic regions (37,(49)(50)(51)(52)(53)(54)(55). Moreover, we annotated genetic variants falling into gene bodies, including deep intronic variants, with AbSplice-DNA, a tool predicting variants causing aberrant splicing (35). On the RNA-seq data, we used OUTRIDER on a total of 12,966 protein-coding genes commonly expressed across the dataset to call high or low expression outliers, and FRASER to call splicing outliers (39,41).…”

“…Using FRASER, which accurately captured the consequence of this initially discarded structural variant as a splicing outlier, we were now able to recognize its significant transcriptomic impact and rescue it. As a complementary approach to RNA-seq-based splicing outlier calling, we considered AbSplice (35), a recently published algorithm that predicts whether a rare variant causes aberrant splicing. Here, we applied it for the first time to hematologic malignancy samples.…”

“…As matched germline tissues for the tumor samples were unavailable, sex-matched genomic DNA from random healthy donors was used as normal material during tumor-normal variant calling. To reduce germline variants and artifacts from the data, we filtered single nucleotide variants, short insertions, and deletions with the following criteria: To predict the effects of variants on splicing, we applied AbSplice-DNA v1.0.0 (35) to the filtered variants. AbSplice-DNA estimates the probability that a rare variant causes aberrant splicing in a given tissue, which takes the variant and splice annotations, so-called SpliceMaps, as input.…”

“…We generated SpliceMaps for each study group as described previously (35). The variants with an AbSplice-DNA prediction score ≥ 0.2 were classified as splice-affecting variants.…”

“…Dataset: Whole genome sequencing and total RNA sequencing of 3,760 hematologic malignancies spanning 24 different disease entities. Bioinformatic processing: On genomic data, intOGen captures recurrent mutational patterns(37), and AbSplice predicts variants causing aberrant splicing(35).…”