Alexey Kolesnikov

DeepSomatic is a deep learning method for detecting somatic variants from short-read and long-read sequencing data, outperforming existing callers, especially for indels.

This study compares Element Avidity sequencing to Illumina sequencing, finding higher accuracy in mapping and variant calling, especially at lower coverages and in challenging genomic regions.

DeepPolisher improves genome assembly accuracy by reducing errors using a transformer model and phasing method, achieving over 70% indel error reduction.

This paper introduces pangenome-aware DeepVariant, a variant caller that improves accuracy by using a pangenome reference, outperforming linear-reference-based DeepVariant by up to 25.5%.

Severus detects and characterizes somatic structural variations in tumor genomes using long reads, outperforming existing methods in accuracy and revealing clinically relevant rearrangements.

The paper develops a benchmark for detecting low-frequency mosaic variants in the HG002 reference material using high-coverage sequencing data and validates its utility across multiple technologies.

This paper introduces a local haplotype approximation method to improve variant calling accuracy across long-read sequencing platforms without increasing computational overhead.

The paper introduces Minigraph-Cactus, a pangenome pipeline that constructs pangenome graphs directly from whole-genome alignments, demonstrating improved performance in mapping and variant calling across human and Drosophila data.

The paper presents a draft human pangenome reference that improves variant discovery and structural variant detection compared to existing references.

The paper shows recombination between heterologous human acrocentric chromosomes using pangenome data, revealing pseudo-homologous regions linked to Robertsonian translocations.

The paper evaluates variant calling in difficult genomic regions using short- and long-read sequencing, showing improved performance with new methods and multi-technology approaches.

The paper introduces a haplotype-aware variant calling pipeline for nanopore and PacBio HiFi data, achieving high accuracy in variant detection and de novo assembly polishing.

DeepConsensus improves PacBio HiFi read accuracy and assembly quality using a gap-aware transformer model.

DeepTrio improves variant calling in family trios using deep learning, achieving higher accuracy than DeepVariant, especially at lower coverages.

This study improves variant calling by integrating population data into deep learning models, enhancing precision and recall across diverse populations.

The paper provides a dataset of 9 gold-standard samples sequenced on multiple platforms and protocols to benchmark sequencing and analysis methods.

This paper develops a long-read sequencing protocol to improve variant detection and genome assembly in a human genome, achieving high accuracy and contiguity.