Александр Владимирович Медведев

BioToken and BioFM - Biologically-Informed Tokenization Enables Accurate and Efficient Genomic Foundation Models

This paper introduces BioToken and BioFM, biologically-informed genomic models that improve performance and efficiency in genomic tasks by incorporating biological context into tokenization.

Tokenization to Transfer: Do Genomic Foundation Models Learn Good Representations?

This study evaluates seven Genomic Foundation Models across 52 genomic tasks, finding that random baselines are strong and tokenizer choices significantly impact performance.

Efficacy of federated learning on genomic data: a study on the UK Biobank and the 1000 Genomes Project

This study evaluates federated learning's effectiveness on genomic data using UK Biobank and 1000 Genomes Project datasets, showing performance close to centralized models despite data heterogeneity.

Human genotype-to-phenotype predictions: boosting accuracy with nonlinear models

This paper compares nonlinear and linear models for genotype-to-phenotype prediction, showing nonlinear models can be more accurate with rich covariates and ensembling improves accuracy for asthma and hypothyroidism.

GRAPE: Genomic Relatedness Detection Pipeline

GRAPE is a genomic relatedness detection pipeline that accurately classifies kinship degrees using preprocessing, IBD detection, and relationship estimation.