Title:	Deep learning for protein - DNA binding predictions
DNr:	Berzelius-2026-284
Project Type:	LiU Berzelius
Principal Investigator:	Emil Marklund <emil.marklund@scilifelab.se>
Affiliation:	Stockholms universitet
Duration:	2026-11-01 – 2027-05-01
Classification:	10307
Homepage:	https://marklundlab.com/
Keywords:

Abstract

Recognition of nucleic acids by regulatory proteins control cellular processes in all life forms. While structural predictions using deep learning methods like AlphaFold have revolutionized our understanding of sequence dependent molecular structure, we have historically had very limited data on how the genetic code specifies binding rates and affinities for interacting molecules. This has prevented us from quantitatively predicting these parameters at arbitrary locations in the genome, and thus limits our ability to understand the sequence and structure dependent gene-regulatory effect of a given biomolecule. In general, we are currently unable to quantitatively predict molecular recognition and function from a nucleotide or protein sequence. Gaining this quantitative understanding is of general interest in all of molecular biology, and of special interest for transcription factor (TF) - DNA binding. This binding controls gene expression, and can cause a myriad of disease states. For example, more than 50% of disease and trait associated genome wide association study variants are found in cis regulatory regions of the genome, like promoter and enhancers, and transcription factors are mutated in most human cancers. Gaining a more quantitative understanding of transcription factor - DNA binding will thus likely be necessary before we can cure these diseases. To understand the sequence dependence on macromolecular binding we need to train quantitative models on high quality biophysical binding data for many sequence mutants. In my lab, we generate this type of data on massively parallel arrays, through 'high-throughput sequencing'-'fluorescent ligand interaction profiling' (HiTS-FLIP). In these assays, we use a second generation sequencing instrument to do affinity and/or kinetic measurements on hundreds of thousands of DNA sequence mutants at the same time. We have recently generated a dataset of binding rates (ka) unbinding rates (kd) and binding equilibrium constants (KD) for the human transcription factor KLF1 binding to a diverse library of 22,349 DNA sequence variants. In the current project, we will train a structure based deep learning model to understand and predict binding affinities for zinc finger TF - DNA interactions, using the HiTS-FLIP binding data for KLF1. We are also focusing on models trained on in vivo, genome wide accessibility (data: ATAC-seq, model: ChromBPNet) and TF (data: ChiP-seq, model: BPNet) binding data. We are performing in silico experiments with the trained models, changing motif syntax at DNA, to gain mechanistic insights how this changes genome accessibility and TF binding. Example of a previous project where we have used in silico evaluations of deep learning models to answer important questions about TF - DNA binding: 1. J. M. Schaepe, T. Fries, B. R. Doughty, V. Ramalingam, B. B. Liu, O. J. Crocker, G. K. Marinov, M. M. Hinks, E. Marklund†, W. J. Greenleaf†, Thermodynamic principles link in vitro transcription factor affinities to single-molecule chromatin states in cells. Cell. 189, 307–322.e23 (2026). † denotes co-corresponding authours.