Biology subjects

Dunin-Horkawicz, S.

Publications and source records attributed to Dunin-Horkawicz, S..

10 recordsLinked to original sources

bioRxiv2026

Simulating protein evolution with programmable fitness objectives derived from modern protein models

Simulation of protein sequence evolution provides a controlled way to test whether specified evolutionary mechanisms can generate the patterns observed in natural proteins. However, classical simulation largely relies on fixed substitution rules, whereas existing fitness-based simulations often define fitness using a single proxy and thus do not sufficiently capture the complex nature of constraints imposed by selection. Here, we simulate protein evolution through amino acid substitutions under programmable fitness objectives derived from individual or combined protein models, including protein language models, inverse folding models, physics-based energy functions, and alignment-based statistical models. Evaluation of simulations revealed that different fitness proxies impose distinct trade-offs between evolutionary realism and molecular plausibility, motivating the use of combined objectives to redistribute selective pressure across complementary constraints. We then show that programmable fitness objectives can encode mechanism-specific biological contexts. Incorporating oligomeric structural context recovered interface-specific evolutionary constraints in homodimeric proteins, while multi-conformation objectives better reproduced natural sequence variation in metamorphic proteins and provided interpretable estimates of how alternative native folds contribute to evolutionary constraint. Finally, we explore a differentiable extension of this framework, in which structural objectives are backpropagated through AlphaFold2 to guide sequence evolution. Under this formulation, evolved sequences showed favorable shifts in several molecular properties relative to those from discrete, stepwise mutation-selection simulations while retaining key evolutionary signatures.

evolutionary biology↗

bioRxiv2026

Beyond profiles: supervised repeat annotation using protein embeddings

Repeated sequence motifs give rise to diverse protein structures and functions, yet their detection is often challenging due to weak sequence similarity between repeat units. Most sensitive approaches rely on homology information directly represented by alignments or derived profiles, thereby limiting flexibility and scalability. Here, we introduce TREAD (Transfer learning-based REpeat Annotation using Protein EmbeDdings), a supervised framework that reformulates repeat detection as a residue-level annotation problem and operates directly on embeddings from protein language models. Instead of constructing explicit probabilistic profiles, TREAD learns repeat-specific features implicitly, enabling residue-resolved scoring and flexible repeat segment extraction. Across complementary benchmarks based on RepeatsDB and Pfam, TREAD consistently matches or outperforms the widely used profile-based tool HMMER, particularly in low-data and high-divergence settings. The model exhibits robustness to score thresholding and demonstrates better generalization to independent test sets, including those containing remote homologs. To illustrate the practical utility of TREAD, we applied it to survey {beta}-propeller proteins across the AlphaFold Database and representative proteomes to generate a comprehensive census of this fold. This analysis highlights extensive propeller diversity, identifies lineage-specific expansion patterns across the tree of life, and suggests previously unrecognized relationships between propellers and other repeat folds. Together, TREAD provides a flexible and scalable alternative to profile-based repeat annotation and establishes a general motif-centric framework for protein sequence annotation.

bioinformatics↗

bioRxiv2026

Conformational diversity in poly-HAMP arrays and its implications for signal transduction

Prokaryotic transmembrane receptors are built around a helical coiled-coil backbone, with specialized sensory, modulatory, and effector domains arranged along its length. The modulatory HAMP domain forms a four-helix parallel coiled coil that is structurally integrated into this backbone, typically connecting transmembrane segments with downstream cytosolic domains. In many systems, HAMP domains have been shown to transduce signals through axial rotation of their helices; however, it is not clear how broadly applicable this mechanism is. Here, we describe two families of soluble chemoreceptors and sensory kinases that contain long arrays of concatenated HAMP domains, which we term poly-HAMP. Although their poly-HAMP arrays have clearly evolved independently, both families share many sequence features consistent with convergence on a similar functional system. We determined the crystal structures of 4-HAMP and 6-HAMP segments from the poly-HAMP array of histidine kinase HskS of Myxococcus xanthus, revealing unusually tight packing between adjacent domains and conformational patterns compatible with the rotational signaling model. To assess whether these features are general and to define the broader conformational landscape of poly-HAMP arrays, we computed AlphaFold2 models of over 200 chemoreceptor- and kinase-associated arrays. The models were consistent with the HskS structures, yet revealed distinct preferences in the two array types: HAMP domains of chemoreceptor arrays were consistently predicted to adopt stable conformations along their lengths, whereas those of kinase arrays were predicted to be biased toward less favorable conformations. By modeling HAMP domains from kinase arrays in isolation from the neighboring domains, we show that they adopt alternative, stable conformations that are related to the array-embedded forms through axial helix rotation. Taken together, our results suggest that, despite their independent origins and structural diversity, HAMP-containing systems ranging from the poly-HAMP arrays studied here through multi-HAMP architectures to canonical single-HAMP receptors may have converged on the same conserved mode of signal transduction that involves axial helix rotation.

molecular biology↗

bioRxiv2025

pLAST - a tool for rapid comparison and classification of bacterial plasmid sequences

MotivationThe increasing number of fully sequenced bacterial plasmids being annotated and cataloged has prompted the development of computational tools for comparing and classifying them. Existing approaches typically compare full-length DNA sequences (e.g., Mash, BLASTn) or translated open reading frames (ORFs) (e.g., BLASTp, DIAMOND), with plasmid-level scores obtained by aggregating ORF-to-ORF similarities; however, they are either restricted to closely related plasmids or become computationally demanding in large-scale analyses. ResultsWe describe pLAST (plasmid Language Analysis and Search Tool), a plasmid-search tool built using word2vec representations capturing both ORF-to-ORF similarity and gene neighborhood conservation. Benchmarks indicate that pLAST outperforms both DNA- and ORF-based methods in identifying functionally similar plasmids and, compared to the widely used Mash, it achieves 37%, 30%, and 13% improvements in detecting shared mating-pair formation (MPF) system, relaxase, and oriT types, respectively. This performance scales to datasets comprising thousands of sequences, as exemplified by clustering analyses of [~]56,000 plasmids, which reveal expected functional groups. Beyond global similarity, pLAST also returns per-ORF plasmid-plasmid alignments, enabling detection of shared functional modules. Availability and implementationpLAST is freely accessible as a web server at https://plast.lbs.cent.uw.edu.pl/ and available as a Python module along with a precomputed database at https://github.com/labstructbioinf/pLAST for customized analysis.

bioinformatics↗

bioRxiv2024

Exploiting protein language model sequence representations for repeat detection

Duplication is an essential evolutionary mechanism that operates at the scale of chromosomes, large chunks of DNA sequences, genes, protein domains, and shorter motifs. The study of duplication is central to understanding protein evolution, but the detection of repetitive sequence patterns is often challenging due to decreasing similarity between internal repeats resulting from long-term divergence. The most sensitive sequence-based repeat detection method, HHrepID, relies on the construction of multiple sequence alignments (MSAs) to enhance homology signals and thus facilitate the detection of very ancient duplications. However, such an alignment-based approach is slow and limits the ability to perform large-scale scans. Recent advances in protein representation learning have introduced sequence embeddings extracted from protein language models as a powerful and much faster alternative to MSAs. Protein sequence representations have been shown to be effective in homology detection, as exemplified by software such as our recently developed pLM-BLAST. In this study, we implement pLM-Repeat, a pipeline built upon pLM-BLAST, to identify repeats encoded in sequence embeddings. pLM-Repeat achieves comparable sensitivity to HHrepID in detecting the presence of repeats, while predicting many more repeat units and providing significantly better run times. We also trained an auxiliary neural network, DeepRepeat, to detect domains with patterns similar to well-characterized repeat folds to support rapid filtering. Using our newly developed tools, we scanned the AFDB90v4 database and identified a collection of novel and undescribed repeat domains.

bioinformatics↗

bioRxiv2024

Applicability of AlphaFold2 in the modelling of coiled-coil domains

Coiled coils are a common protein structural motif involved in cellular functions ranging from mediating protein-protein interactions to facilitating processes such as signal transduction or regulation of gene expression. They are formed by two or more alpha helices that wind around a central axis to form a buried hydrophobic core. Various forms of coiled-coil bundles have been reported, each characterized by the number, orientation, and degree of winding of the constituent helices. This variability is underpinned by short sequence repeats that form coiled coils and whose properties determine both their overall topology and the local geometry of the hydrophobic core. The strikingly repetitive sequence has enabled the development of accurate sequence-based coiled-coil prediction methods; however, the modeling of coiled-coil domains remains a challenging task. In this work, we evaluated the accuracy of AlphaFold2 in modeling coiled-coil domains, both in modeling local geometry and in predicting global topological properties. Furthermore, we show that the prediction of the oligomeric state of coiled-coil bundles can be achieved by using the internal representations of AlphaFold2, with a performance better than any previous state-of-the-art method (code available at https://github.com/labstructbioinf/dc2_oligo).

bioinformatics↗

bioRxiv2023

AlphaFold2 captures the conformational landscape of the HAMP signaling domain

In this study, we present a conformational landscape of 5000 AlphaFold2 models of the HAMP domain, a short helical bundle that transduces signals from sensors to effectors in two-component signaling proteins such as sensory histidine kinases and chemoreceptors. The landscape reveals the conformational variability of the HAMP domain, including rotations, shifts, displacements, and tilts of helices, many combinations of which have not been observed in experimental structures. HAMP domains belonging to a single family tend to occupy a defined region of the landscape, even when their sequence similarity is low, suggesting that individual HAMP families have evolved to operate in a specific conformational range. The functional importance of this structural conservation is illustrated by poly-HAMP arrays, in which HAMP domains from families with opposite conformational preferences alternate, consistent with the rotational model of signal transduction. The only poly-HAMP arrays that violate this rule are predicted to be of recent evolutionary origin and structurally unstable. Finally, we identify a family of HAMP domains that are likely to be dynamic due to the presence of a conserved pi-helical bulge. All code associated with this work, including a tool for rapid sequence-based prediction of the rotational state in HAMP domains, is deposited at https://github.com/labstructbioinf/HAMPpred.

bioinformatics↗

bioRxiv2022

Protein modularity in phages is extensive and associated with functions linked to core replication machinery and host tropism determinants

Biological modularity enhances evolutionary adaptability by allowing rearrangement of functional components. One striking example are bacterial viruses (phages). They exhibit extensive genomic modularity by being built of independent functional modules that evolve separately and combine in various ways, making them astoundingly diverse. While multiple studies have investigated genomic modularity in phages, less attention has been given to protein modularity--proteins having distinct building blocks or domains that can evolve and recombine, enhancing functional and genetic diversity. To better understand the impact of protein modularity on viral evolution, we quantified it by detecting instances of domain mosaicism, defined as a homologous fragment sharing between two otherwise unrelated proteins. We used highly sensitive homology detection to quantify domain mosaicism between pairs of 133,574 representative phage proteins and to understand its relationship with functional diversity in phage genomes. We found that diverse functional classes often shared homologous domains. This phenomenon was often linked to protein modularity, particularly in receptor-binding proteins, endolysins and DNA polymerases. We also identified multiple instances of recent diversification via exchange and gain/loss of domains in receptor-binding proteins, neck passage structures, endolysins and some members of the core replication machinery. Diversification via protein fragment exchange often transcended distant taxonomic and ecological borders. We argue that the ongoing diversification via shuffling of protein domains associated with those functions is reflective of co-evolutionary arms race and the resulting diversifying selection to overcome multiple mechanisms of bacterial resistance against phages.

evolutionary biology↗

bioRxiv2022

pLM-BLAST - distant homology detection based on direct comparison of sequence representations from protein language models

MotivationThe detection of homology through sequence comparison is a typical first step in the study of protein function and evolution. In this work, we explore the applicability of protein language models to this task. ResultsWe introduce pLM-BLAST, a tool inspired by BLAST, that detects distant homology by comparing single-sequence representations (embeddings) derived from a protein language model, ProtT5. Our benchmarks reveal that pLM-BLAST maintains a level of accuracy on par with HHsearch for both highly similar sequences (with over 50% identity) and markedly divergent sequences (with less than 30% identity), while being significantly faster. Additionally, pLM-BLAST stands out among other embedding-based tools due to its ability to compute local alignments. We show that these local alignments, produced by pLM-BLAST, often connect highly divergent proteins, thereby highlighting its potential to uncover previously undiscovered homologous relationships and improve protein annotation. Availability and ImplementationpLM-BLAST is accessible via the MPI Bioinformatics Toolkit as a web server for searching precomputed databases (https://toolkit.tuebingen.mpg.de/tools/plmblast). It is also available as a standalone tool for building custom databases and performing batch searches (https://github.com/labstructbioinf/pLM-BLAST).

bioinformatics↗

bioRxiv2021

Graph neural networks and sequence embeddings enable the prediction and design of the cofactor specificity of Rossmann fold proteins

The Rossmann fold enzymes are involved in essential biochemical pathways such as nucleotide and amino acid metabolism. Their functioning relies on interaction with cofactors, small nucleoside-based compounds specifically recognized by a conserved {beta}{beta} motif shared by all Rossmann fold proteins. While Rossmann methyltransferases recognize only a single cofactor type, the S-Adenosylmethionine (SAM), the oxidoreductases, depending on the family, bind nicotinamide (NAD, NADP) or flavin-based (FAD) cofactors. In this study, we show that despite its short length, the {beta}{beta} motif unambiguously defines the specificity towards the cofactor. Following this observation, we trained two complementary deep learning models for the prediction of the cofactor specificity based on the sequence and structural features of the {beta}{beta} motif. A benchmark on two independent test sets, one containing {beta}{beta} motifs bearing no resemblance to those of the training set, and the other comprising 38 experimentally confirmed cases of rational design of the cofactor specificity, revealed the nearly perfect performance of the two methods. The Rossmann-toolbox protocols can be accessed via the webserver at https://lbs.cent.uw.edu.pl/rossmann-toolbox and are available as a Python package at https://github.com/labstructbioinf/rossmann-toolbox. Key pointsO_LIThe Rossmann fold encompasses a multitude of diverse enzymes involved in most of the essential cellular pathways C_LIO_LIProteins belonging to the Rossmann fold co-evolved with their nucleoside-based cofactors and require them for the functioning C_LIO_LIManipulating the cofactor specificity is an important step in the process of enzyme engineering C_LIO_LIWe developed an end-to-end pipeline for the prediction and design of the cofactor specificity of the Rossmann fold proteins C_LIO_LIOwing to the utilization of deep learning approaches the pipeline achieved nearly perfect accuracy C_LI

bioinformatics↗