Biology subjects

Dima, R. I.

Publications and source records attributed to Dima, R. I..

12 recordsLinked to original sources

bioRxiv2026

Building from the ground up: augmentation strategies to improve protein language model performance on protein sets

An important underlying requirement for deep learning models is the availability of large datasets, however, in sciences such as chemistry and biology, generating experimental data of sufficient size represents a significant bottleneck. In this work, we probed the extent and nature of limited input data, characterize how several protein language models (pLMs) learned on suboptimal datasets, and investigated augmentation approaches to improve predictive model performance. We found that narrowing the distribution of protein-protein complexes in a limited dataset permits an order-of-magnitude decrease in training dataset size, if a model is only expected to generalize to in-distribution data. Similar results were found in active learning (AL), where both in- and out-of-distribution Matthews Correlation Coefficient (MCC) showed a small but consistent increase upon augmentation of the training set with complexes produced by ESM-3, a generative pLM.

biophysics↗

bioRxiv2026

Probing Mechanisms of Allosteric Regulation in AAA+ ATPases for Microtubule Severing and Protein Disaggregation

Ring-like AAA+ (ATPases Associated with diverse cellular Activities) biological machines mediate protein remodeling to assist a broad range of essential cellular functions. The nucleotide-dependent remodeling action involves intra- and inter-ring allosteric communication to generate mechanical force applied onto the substrate by a set of loops that protrude into the central channel. In this study, we probe these allosteric mechanisms through a comparative study of the katanin, a microtubule severing protein including a clade 3 AAA domain, and the double-ring ClpB, a protein disaggregase including both a clade 3 and a clade 5 AAA domain. Our molecular dynamics simulations, combined with machine learning and bioinformatic analysis, reveal both similar mechanisms involving the clade 3 domain and ClpB-specific ones involving communication with the clade 5 domain. We find that both nucleotide and substrate polypeptide binding restrict the conformational landscape sampled by katanin and ClpB, with ligand-specific conformations observed in the latter case. Allosteric contributions of secondary structure elements, ranked by using SHapley Additive exPlanations analysis in machine learning approaches and binary classification of features in ligand states, highlight the important role of regions adjacent to the nucleotide-binding site and the pore loops. Amino acid-level analysis of the allosteric paths reveals that intra-ring cooperativity modulates long-distance communication within the AAA+ protomers.

biophysics↗

bioRxiv2026

Distinct Tubulin C-Terminal Tails Control the Efficiency of a Microtubule Severing Machine

Microtubule severing enzymes of the AAA+ family are essential regulators of cytoskeletal remodeling, extracting subunits from the microtubule lattice through ATP-driven conformational changes. Among them, katanin assembles into hexameric structures and binds the negatively charged carboxy-terminal tails (CTTs) of tubulin through its central pore. Experimental studies have shown that different tubulin CTT isotypes can act as either inhibitors or non-inhibitors of katanin-mediated severing, with increased CTT hydrophobicity associated with reduced inhibition. However, the molecular basis underlying this selective behavior remains poorly understood. Here, we employed molecular dynamics simulations, and quantitative analysis combining principal component analysis, clustering, and distance distribution analysis, to investigate how natural tubulin CTTs (beta5, beta4b, and beta3), and engineered CTTs (beta5-A+Y, beta5-cterm, beta5-midpoint, and poly-E) influence katanin structure and dynamics in spiral and ring conformations. Our results show that inhibitory CTTs form stronger interactions with the terminal protomers and increase flexibility in the inner protomers, resulting in coordinated motions associated with pore narrowing. In contrast, non-inhibitory CTTs preferentially interact with inner protomers, disrupt interprotomer coordination, weakening the collective grip of the hexamer on the substrate. Analyses of the engineered constructs further revealed that inhibitory behavior is governed primarily by the spatial distribution of acidic residues rather than the overall charge. Additionally, we identified species-specific responses of different CTT isotypes toward the katanin ring state. Together, these findings provide molecular-level insight into how tubulin CTT sequence organization regulates substrate recognition, pore dynamics, and severing efficiency, leading to predictive design of CTTs with a desired action on katanin.

biophysics↗

bioRxiv2026

FlexENN: A Graph Neural Network for Binding Energy Prediction of Globular and Intrinsically Disordered Proteins

Intrinsically disordered proteins (IDPs) drive a large fraction of cellular signaling, transcription and assembly through interfaces that lack a single defined geometry, making the prediction of their binding energies beyond the reach of methods calibrated on the rigid complementarity of folded complexes. Here we introduce an adaptive message passing graph neural network, FlexENN, that predicts binding energies across the full structural spectrum, from folded domains to complexes with one disordered partner. The FlexENN architecture constructs a local interface graph in which each residue node integrates relative geometry, dynamic descriptors, and sequence information, allowing the network to infer conformational fuzziness from a single structure. Benchmarked across folded complexes and IDP systems, including tubulin-tubulin interfaces within microtubules (MTs) and MTs in complex with microtubule associated proteins (MAPs), with reference binding affinity values extracted from our experiments, the model retrieves physically meaningful energetic profiles for complexes where rigid-body methods fail. More broadly, this work delivers an accurate, structure-aware approach to predicting binding free energies for the broad class of disordered-mediated complexes formed by IDPs.

biophysics↗

bioRxiv2024

Structural and functional role of residues undergoing hereditary spastic paraplegias-linked mutations: insights from a simulation of the spiral to ring transition in katanin

Several dozen mutations in the large human spastin enzyme assembly have been associated with mobility impairment in hereditary spastic paraplegias. Some of them impact the structural determinants of two functional conformations of the protein, spiral and ring. Here we investigate the possible effect of the mutations on the transition between the two conformations, which is essential for the enzymatic function. By performing a variety of molecular simulations (including metadynamics) on the closely related protein katanin, we suggest that about one fourth of the known disease-associated mutations affect the transition and/or the stability of a previously unrecognized intermediate. The protocol used here can be applied to the study of conformational changes in other large biomolecular complexes. Significance StatementBy combining molecular dynamics and enhanced sampling simulations with custom-designed collective variables, we explore transient intermediate states in a large protein assembly, the severing enzyme katanin. This enzyme plays a critical role in neuronal microtubule remodeling. By targeting the transition between the spiral and ring conformations of the protein, we reveal mechanisms associated with neurodegenerative mutations that impair the function of spastin, a structurally and functionally similar enzyme. These findings provide insights that may inform future therapeutic strategies. They also expand our ability to simulate molecular processes relevant to health and disease, as our approach can be readily applied to study structural changes in other complex biological systems.

biophysics↗

bioRxiv2024

A major disease-related point mutation in spastin alters dramatically the dynamics and allostery of the motor

Spastin is a microtubule-severing AAA+ ATPase that is highly expressed in neu-ronal cells and plays a crucial role in axonal growth, branching, and regeneration. This machine oligomerizes into hexamers in the presence of ATP and the microtubule carboxy-terminal tails (CTTs). Conformational changes in spastin hexamers, pow-ered by ATP hydrolysis, apply forces on the microtubule, ultimately leading to the severing of the filament. Mutations disrupt the normal function of spastin, impair-ing its ability to sever microtubules effectively and leading to abnormal microtubule dynamics in neurons characteristic for the set of neurodegenerative disorders called hereditary spastic paraplegias (HSP). Experimental studies have identified the HSP-related R591S (Drosophila melanogaster numbering) mutation as playing a crucial role in spastin. Given its significant role in HSP, we employed a combination of molecular dynamics simulations with machine learning and graph network based approaches to identify and quantify the perturbations caused by the R591S HSP mutation on spastins dynamics and allostery with functional implications. We found that the functional hex-amer, upon the HSP-related mutation, loses the ability to execute the primary motion associated with the severing action. The study of allosteric changes upon the mutation showed that the regions that are most perturbed are those involved in the formation of the inter-protomer contacts. The mutation induces rigidity in the allosteric networks of the motor making it more likely to experience loss of function as any applied per-turbations could not be easily dissipated by passing through a variety of alternative paths as in the wild-type (WT) spastin.

biophysics↗

bioRxiv2024

Machine Learning Approaches in Label-Free Small Extracellular Vesicles Analysis with Surface-Enhanced Raman Scattering (SERS) for Cancer Diagnostics

Early diagnosis remains of pivotal importance in reducing patient morbidity and mortality in cancer. To this end, liquid biopsy is emerging as a tool to perform broad cancer screenings. Small extracellular vesicles (sEVs), also called exosomes, found in bodily fluids can serve as important cancer biomarkers in these screenings. Our group has recently developed a label-free electrokinetic microchip to purify sEVs from blood. Herein, we demonstrate the feasibility to integrate this approach with surface-enhanced Raman scattering (SERS) analysis. SERS can be used to characterized extracted sEVs through their vibrational fingerprint that changes depending on the origin of sEVs. While these changes are not easily identified in spectra, they can be modeled with machine learning (ML) approaches. Common ML approaches in the field of spectral analysis use dimensionality reduction method that often function as a black box. To avoid this pitfall, we used Shapley additive explanations (SHAP) is a type of explainable AI (XAI) that bridges ML models and human comprehension by calculating the specific contribution of individual features to a models predictions, directly correlating model/decisions with the original data. Using these approaches we demonstrated a proof-of-concept model predictive of cancer from isolated sEVs, integrating the electrokinetic device and SERS. This work explores the use of explainable AI to perform diagnostic analysis on complex SERS data of clinical samples, while reporting interpretable biochemical information. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=74 SRC="FIGDIR/small/581099v2_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@1cbcb8aorg.highwire.dtl.DTLVardef@9ffc00org.highwire.dtl.DTLVardef@15999a9org.highwire.dtl.DTLVardef@1775726_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2023

Searching for Structure: Characterizing the Protein Conformational Landscape with Clustering-based Algorithms

The identification and characterization of the main conformations from a protein population is a challenging, inherently high-dimensional problem. We introduce the Secondary sTructural Ensembles with machine LeArning (StELa) double clustering method, which clusters protein structures based on the underlying Ramachandran plot. Our approach takes advantage of the relationship between the phi and psi dihedral angles in a protein backbone and the secondary structure of the protein. The classification of states as vectors composed of the clusters indices arising naturally from the Ramachandran plot, followed by the hierarchical clustering of the vectors, enables the identification of the minima from the corresponding free energy landscape (FEL) by lifting the high structure degeneracy found with existing approaches such as the RMSD-based clustering GROMOS. We compare the performance of StELa with not only GROMOS but also with CATS, the combinatorial averaged transient structure clustering method based on distributions of the phi and psi dihedral angle coordinates. Using ensembles of conformations from molecular dynamics (MD) simulations of either intrinsically disordered proteins (IDPs) of various lengths (tau protein fragments) or from local structures from a globular protein, we show that StELa is the only clustering method that identifies nearly all the minima from the corresponding FELs. In contrast, GROMOS yields a large number of clusters that cover the entire FEL and CATS, even with an additional clustering step, is unable to sample well the FEL for long IDPs and for fragments from globular proteins as it misses important minima. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/557631v1_ufig1.gif" ALT="Figure 1"> View larger version (36K): org.highwire.dtl.DTLVardef@634f90org.highwire.dtl.DTLVardef@1fca966org.highwire.dtl.DTLVardef@d59bd9org.highwire.dtl.DTLVardef@1eadb94_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2022

Microtubule severing enzymes oligomerization and allostery: a tale of two domains

Severing proteins are nanomachines from the AAA+ (ATPases associated with various cellular activities) superfamily whose function is to remodel the largest cellular filaments, microtubules. The standard AAA+ machines adopt hexameric ring structures for functional reasons, while being primarily monomeric in the absence of the nucleotide. Both major severing proteins, katanin and spastin, are believed to follow this trend. However, studies proposed that they populate lower-order oligomers in the presence of co-factors, which are functionally relevant. Our simulations show that the preferred oligomeric assembly is dependent on the binding partners, and on the type of severing protein. Essential dynamics analysis predicts that the stability of an oligomer is dependent on the strength of the interface between the helical bundle domain (HBD) of a monomer and the convex face of the nucleotide binding domain (NBD) of a neighboring monomer. Hot spots analysis found that the region consisting of the HBD tip and the C-terminal (CT) helix is the only common element between the allosteric networks responding to nucleotide, substrate, and inter-monomer binding. Clustering analysis indicates the existence of multiple pathways for the transition between the secondary structure of the HBD tip in monomers and the structure(s) it adopts in oligomers.

biophysics↗

bioRxiv2022

Exploring the effect of mechanical anisotropy of protein structures in the unfoldase mechanism of AAA+ molecular machines

Essential cellular processes of microtubule disassembly and protein degradation, which span lengths from tens of m to nm, are mediated by specialized molecular machines with similar hexameric structure and function. Our molecular simulations at atomistic and coarse-grained scales show that both the microtubule severing protein spastin and the caseinolytic protease ClpY, accomplish spectacular unfolding of their diverse substrates, a microtubule lattice and dihydrofolate reductase (DHFR), by taking advantage of mechanical anisotropy in these proteins. By considering wild-type and variants of DHFR, we found that optimal ClpY-mediated action probes favorable orientations of the substrate relative to the machine. Unfolding of wild-type DHFR involves strong mechanical interfaces near each terminal and occurs along branched pathways, whereas unfolding of DHFR variants involves softer mechanical interfaces and occurs through single pathways, but translocation hindrance can arise from internal mechanical resistance. For spastin, optimum severing action initiated by pulling on a tubulin subunit is achieved through the orientation of the machine versus the substrate (microtubule lattice). Moreover, changes in the strength of the interactions between spastin and a microtubule filament, which can be driven by the tubulin code, lead to drastically different outcomes for the integrity of the hexameric structure of the machine.

biophysics↗

bioRxiv2020

Factors Underlying Asymmetric Dynamics of Disaggregase and Microtubule Severing AAA+ Machines

Disaggregation and microtubule-severing nanomachines from the AAA+ (ATPases associated with various cellular activities) superfamily assemble into ring-shaped hexamers that enable protein remodeling by coupling large-scale conformational changes with application of mechanical forces within a central pore by loops protruding within the pore. We probed these motions and intra-ring interactions that support them by performing extensive explicit solvent molecular dynamics simulations of single-ring severing proteins and the double-ring disaggregase ClpB. Simulations reveal that dynamic stability of hexamers of severing proteins and of the nucleotide binding domain 1 (NBD1) ring of ClpB, which belong to the same clade, involves a network of salt bridges that connect conserved motifs of central PL1 loops of the hexamer. Clustering analysis of ClpB highlights correlated motions of domains of neighboring protomers supporting strong inter-protomer collaboration. Severing proteins have weaker inter-protomer coupling and stronger intra-protomer stabilization through salt bridges formed between PL2 and PL3 loops. Distinct mechanisms are identified in the NBD2 ring of ClpB involving weaker inter-protomer coupling through salt bridges formed by non-canonical loops and stronger intra-protomer coupling. Pore width fluctuations associated with the PL1 constriction in the spiral states, in the presence of a substrate peptide, highlight stark differences between narrowing of channels of severing proteins and widening of the NBD1 ring of ClpB. This indicates divergent substrate processing mechanisms of remodeling and translocation by ClpB and substrate tail-end gripping and possible wedging on microtubule lattice by severing enzymes. Relaxation dynamics of the distance between the PL1 loops and the centers of mass of protomers reveals observation-time-dependent dynamics, leading to predicted relaxation times of tens of microseconds on millisecond experimental timescales. For ClpB the predicted relaxation time is in excellent agreement with the extracted time from smFRET experiments.

biophysics↗

bioRxiv2019

Molecular Investigations into the Unfoldase Action of Severing Enzymes on Microtubules

Microtubules associated proteins regulate the dynamic behavior of microtubules during cellular processes. Microtubule severing enzymes are the associated proteins which destabilize microtubules by removing subunits from the lattice. One model for how severing enzymes remove tubulin dimers from the microtubule lattice is by unfolding its subunits through pulling on the carboxy-terminal tails of tubulin dimers. This model stems from the fact that severing enzymes are AAA+ unfoldases. To test this mechanism, we apply pulling forces on the carboxy-terminal regions of microtubule subunits using coarse grained molecular simulations. In our simulations we used different microtubule lattices and concentrations of severing enzymes. We compare our simulation results with data from in-vitro severing assays and find that the experimental data is best fit by a model of cooperative removal of protofilament fragments by severing enzymes which depends on the severing enzyme concentration and placement on the microtubule lattice.

biophysics↗