deepvariant

DeepVariant is a deep learning-based variant caller that processes aligned genomic reads to produce pileup image tensors, which are then classified using a convolutional neural network and output as VCF or gVCF files. Its primary use case is germline variant calling in diploid organisms, capable of handling diverse input data sources such as Illumina, PacBio, and Oxford Nanopore technologies. Notable features include support for hybrid datasets, pangenome-aware calling, and the extension to trio variant calling through the DeepTrio module, facilitating nuanced analysis of genomic variants in family-based studies.