Physics and Evolution of Biological Macromolecules




Working at the interface between physics and biology, we develop physical models and computational approaches for modelling structures of biomolecules and for predicting their functional characteristics. The research topics in the group include, but are not limited to, protein allostery, biophysics of chromatin and epigenetic regulation, evolution of protein function, evolutionary-inspired design of protein function, and molecular mechanisms of adaptation.


Our work on basic mechanisms and practical implications of allostery, have continuously propelled us to be among the world leaders in the cutting-edge research on allostery. As an indicator of the group’s achievements, the Special Issue “Allostery: From Mechanisms to Therapies” of the Journal of Molecular Biology published this year was edited by Igor Berezovsky and Ruth Nussinov [1]. The Issue contains a collection of works, covering different aspects of allostery on multiple scales, from intramolecular regulation of protein function to intercellular signalling.

Figure 1. Multiscale allostery in action: gearbox of molecular regulation and cellular signalling. The cover illustration of the Journal of Molecular Biology Special Issue “Allostery: From Mechanisms to Therapies” designed by Wei-Ven Tee and Igor Berezovsky.

Continuing our broad studies on fundamental and practical aspects on allostery [2-6] we contributed to this Issue two original works: (i) one considers the role of evolutionary conservatism and diversity in establishing of allosteric communication in different types of folds/domains and their multidomain/chain assemblies [7]; (ii) another project explores features of allosteric ligands, sites, and interactions with the aim to develop computational models for allosteric drug design [8]. More specifically, in the first work we learn directly from nature how various major folds provide structural platforms for allosteric regulation of many enzymatic and signalling functions. We obtained a picture of conserved allosteric communication characteristic in different fold types, including α/ß and ß-barrels, ß-propellers, Ig-like fold, ankyrin and α/ß leucine-rich repeat proteins, modifications of the structuredriven signalling patterns via sequence-determined divergence to specific functions, as well as the emergence and potential diversification of allosteric regulation in multi-domain proteins and oligomeric assemblies. Our observations facilitate the engineering and de novo design of proteins [9] with allosterically regulated functions, including development of therapeutic biologics.

Figure 2. Archetypal patterns of allosteric communication and their potential alterations in different folds. Inner circle: Conserved patterns of allosteric communication characteristic for various major folds. Outer circle: Diversity of allosteric phenomena and potential for their engineering in different proteins and oligomeric assemblies.

The other contribution “Learning about allosteric drugs and ways to design them” [8] seeks to answer the precision medicine quest on individually targeting and selective drugs with the ability to target the so-called difficult or undruggable targets. We argue that prospected allosteric drugs possess several traits satisfying requirements of personalized therapies, including non-competitiveness, spatiotemporal specificity, as well as pathway and substrate specificity. Complemented by the modulatory rather than on/off mode of action with longlasting changes and ceiling effects, allosteric effectors allow to circumvent or at least to significantly alleviate off-target toxicity and drug resistance typical for traditional orthosteric medicines, also providing an opportunity to implement personalized, accurately tuned drug dosing. We propose a generic protocol for computational design of allosteric effectors, enabling also the allosteric tuning of biologics, in obtaining allosteric control over protein functions.


We continue to work on our web-resources for the analysis of allostery, the AlloSigMA server [10] and the AlloMAPS database [11]. Recently, we released the AlloMAPS 2 - an update of the Allosteric Mutation Analysis and Polymorphism of Signalling database, which contains data on allosteric communication obtained for predicted structures in the AlphaFold database (AFDB) and trRosetta-predicted Pfam domains. The data update contains Allosteric Signalling Maps (ASMs) and Allosteric Probing Maps (APMs) quantifying allosteric effects of mutations and of small probe binding, respectively. To ensure quality of the ASMs and APMs, we performed careful and accurate selection of protein sets from high-quality predicted structures in both databases for each organism/structure. We show that many lowquality AlphaFold predictions may require special treatments. For example, AlphaFold predictions of multi-chain and/or oligomeric proteins should be performed first for individual chains/protomers before assembly into the final structure. The fingerprints of Allosteric Signalling and Probing Maps (ASMs/ APMs) provided here for newly predicted high-quality structures allow to expand the analysis of allosteric communication from the effects of ligand binding and/or mutations to engineering and design of allosteric signalling in newly predicted structures, as well as to target them with newly developed allosteric effectors. Specifically, the ASM fingerprints can be instrumental in diagnostics of broad pathologies, helping to predict allosteric effects of single/multiple mutations, and analyzing their role in the expansion of cancer landscapes and involvements in other diseases. We also anticipate that the exhaustive character of allosteric signalling and probing fingerprints will be useful in future developments of corresponding machine learning applications.

Figure 3. The logogram of the dataset for predicted structures in the AlphaFold database.


We continue to develop a computational framework for reconstructing whole-genome chromatin ensemble from Hi-C data [12], and for exploring how structures and dynamics of chromatin drive genome expression. The current twostage protocol includes Markov state modelling (MSM) for reconstructing the structural hierarchy of chromatin organization with partitioning and effective interactions, and the stochastic embedding procedure (SEP) for obtaining the 3D ensemble reconstruction.


Figure 4. The flowchart of the computational framework for the analysis of chromatin hierarchy and the 3D whole-genome chromatin reconstruction.


 Senior Principal Investigator BEREZOVSKY Igor N.   |    [View Bio]   
 Senior Scientist II  SU Tran-to Chinh 
 Scientist  TEE Wei Ven
 Scientist  DONG Bingxue
 Research Officer KRITHIKA Subramani
 PhD Student AMANGELDINA Aidana
 Research Officer Raechell

Selected Publications