Harmen J. Bussemaker

The paper develops a method to predict how mutations in transcription factors affect their DNA binding specificity using biophysically interpretable machine learning.

The paper develops a sequence-to-affinity model for SH2 domains using multi-round peptide binding assays and free-energy regression to predict binding specificity and variant effects.

The paper benchmarks DNA binding affinity models using allele-specific ChIP-seq data to predict allelic imbalances and improve TF binding models.

The paper identifies genetic variants affecting transcription factor activity by analyzing RNA-seq data from GTEx and using a beta-binomial GLM to map aQTLs.

The paper systematically examines how TP53 binding site architecture affects transcriptional output and develops a model to predict reporter activity.

The paper shows Hox paralogs use different cofactors in distinct domains to activate specific gene regulatory networks in Drosophila legs.

The paper introduces ProBound, a machine learning framework that quantifies biophysical parameters of protein-ligand interactions using NGS data, and couples it with Kd-seq to determine absolute affinities.

The paper identifies 6,262 TF-eQTL interactions across 1,598 genes using GTEx data, linking genetic variants to transcription factor regulation and GWAS loci.

The paper investigates how DNA sequence influences the formation of transcription factor complexes and shows that non-canonical binding sites arise from direct interactions with functionally bound sites.

The paper identifies over 30,000 SNPs affecting regulatory element activity using SuRE reporter technology, linking them to gene regulation and potential human traits.

This study identifies differences in DNA-binding preferences between androgen and glucocorticoid receptors using SELEX-seq and SelexGLM, revealing AR's greater sequence sensitivity and enthalpy-driven binding.

The paper investigates how sequence mismatches affect Cas9 binding affinity and endonuclease activity, finding that these effects are sometimes decoupled and depend on the PAM region.

The paper introduces K-mer Motif Multinomial Mixtures (KMMMs), a Bayesian framework for motif discovery in sequence data, and applies it to DNAse hypersensitive sites and ChIP-seq peaks.