DeepTrio: Variant Calling in Families Using Deep Learning

Abstract

Every human inherits one copy of the genome from their mother and another from their father. Parental inheritance helps us understand the transmission of traits and genetic diseases, which often involve de novo variants and rare recessive alleles. Here we present DeepTrio, which learns to analyze child-mother-father trios from the joint sequence information, without explicit encoding of inheritance priors. DeepTrio learns how to weigh sequencing error, mapping error, and de novo rates and genome context directly from the sequence data. DeepTrio has higher accuracy on both Illumina and PacBio HiFi data when compared to DeepVariant. Improvements are especially pronounced at lower coverages (with 20x DeepTrio roughly equivalent to 30x DeepVariant). As DeepTrio learns directly from data, we also demonstrate extensions to exome calling solely by changing the training data. DeepTrio includes pre-trained models for Illumina WGS, Illumina exome, and PacBio HiFi.

DeNovoCNN: A deep learning approach to de novo variant calling in next generation sequencing data

“…Recently, DeepTrio (7), an extension of DeepVariant (22) was developed. In our study we did not compare our results to those of DeepTrio because it is fundamentally a variant calling algorithm, where in order to identify DNMs, the genotypes of the variants are combined to generate a list of candidate DNMs.…”

Clair3-Trio: high-performance Nanopore long-read variant calling in family trios with Trio-to-Trio deep neural networks

“…We computed the number of Mendelian violation variants in trios using the following steps: (1) merging all trio variants results using BCFtools (v1.12) [20] with the flag “-f PASS -0 -m all”, and (2) computing the number of Mendelian violations via RTG tools (v3.12.1) [21]. We also computed the number of de novo variants in the model’s prediction, where the de novo variants [15] are defined as variants confidently genotyped as 0/1 in the child and as 0/0 or unknown in the parents. Note that the metrics of Precision, Recall, F1-score, and number of de novo variants are constrained in the confidence region, while the number of the Mendelian violations is computed in all sites.…”

A Complete Pedigree-Based Graph Workflow for Rare Candidate Variant Analysis

“…Following 12 alignment, samples in the trio are variant-called, producing a per-sample gVCF genotype called file. A trio-based DeepVariant extension (Poplin et al 2018), Google's DeepTrio (Kolesnikov et al 2021), is used to call variants in this workflow. DeepTrio first generates images based on the alignments between the parent and child reads.…”

“…While pangenome graphs have helped to reduce reference mapping bias, further performance improvements are possible. We introduce VG-Pedigree, a pedigree-aware workflow based on the pangenome-mapping tool of Giraffe (Sirén et al 2021) and the variant-calling tool DeepTrio (Kolesnikov et al 2021) using a specially-trained model for Giraffe-based alignments. We demonstrate mapping and variant calling improvements in both single-nucleotide variants (SNVs) and insertion and deletion (INDEL) variants over those produced by alignments created using BWA-MEM to a linear-reference and Giraffe mapping to a pangenome graph containing data from the 1000 Genomes Project.…”