Biology subjects

Lihan, M.

Publications and source records attributed to Lihan, M..

5 recordsLinked to original sources

bioRxiv2026

Unsupervised protein language models learn patterns of enzyme function

While enormous amounts of sequence information have become available, assignment of sequence to a particular enzymatic function has remained elusive. Here we describe a framework that drives a general protein language model to find a target reaction without specific training, using an initial bridgehead protein. At the heart of this framework is PLM-clust, an algorithm that employs k-means on top of protein language model embeddings to convert sequence space into functional reservoirs of latent space, and samples from these clusters based on accelerated zero-shot scoring. We demonstrate PLM-clust in a recursive discovery process (with enzyme hit rates quickly rising to >90%), segmenting isofunctional reservoirs and exploring them in greater detail. This approach - exemplified for glycosyl hydrolases (a xylanase, >100-fold activity increase) and for imine reductases (IREDs, >100-fold increase in catalytic promiscuity profiles) - reliably brings about novel enzymes that are proficient at the catalytic task at hand, reaching deeply into sequence space with a majority of residues exchanged.

synthetic biology↗

bioRxiv2026

Results of a large scale study of the binding of 50 type IIinhibitors to 348 kinases: The role of protein reorganization

Kinase family proteins constitute the second largest protein class targeted in drug development efforts, most prominently to treat cancer, but also several other diseases associated with kinase dysfunction. In this work we focus on type II kinase inhibitors which bind to the "classical" inactive conformation of the protein kinase catalytic domain where the DFG motif has a "DFG-out" orientation and the activation loop is folded. Many Tyrosine kinases (TKs) exhibit strong binding affinity with a wide spectrum of type II inhibitors while serine/threonine kinases (STKs) often bind more weakly. Recent work suggests this difference is largely due to differences in the folded to extended conformational equilibrium of the activation loop between TKs vs. STKs. The binding affinity of a type II inhibitor to its kinase target can be decomposed into a sum of two contributions: (1) the free energy cost to reorganize the protein from the active to inactive state, and (2) the binding affinity of the type II inhibitor to the inactive kinase conformation. In previous work we used a Potts statistical energy potential based on sequence co-variation to thread sequences over ensembles of active and inactive kinase structures. The threading function was used to estimate the free energy cost to reorganize kinases from the active to classical inactive conformation, and we showed that this estimator is consistent with the results of molecular dynamics free energy simulations for a small set of STKs and TKs. In the current study, we analyze the results of a large-scale study of the binding affinities of 50 type II inhibitors to 348 kinases, of which the results for 16 of the 50 type II inhibitors were reported in an earlier study (the "Davis dataset"); the binding data for the remaining 34 type II inhibitors to the panel of 348 kinases were recently obtained (the "Schrodinger dataset"). We use the Potts statistical energy model to investigate the contribution of protein reorganization to the selectivity of the large kinase panel against the set of 50 type II inhibitors, and find that protein reorganization makes a significant contribution to the selectivity. The AUC of the receiver-operator characteristic curve is [~]0.8. We report the results of an internal "blind test", that shows how Potts threading energies can provide more accurate estimates of kinase selectivity than corresponding predictions using experimental results of small sample size. We discuss why two STK phylogenetic kinase families, STE and CMGC, appear to contain many outliers, and how to improve the ability to predict kinase selectivity with a more complete analysis of the kinase conformational landscape. We compare the performance of Potts threading for predicting binding properties of the large set of (50) Type II inhibitors to 348 kinases, with those of a sequence-based purely machine learning model, DeepDTAGen, a publicly available machine learning model that was trained on the complete Davis dataset, including both Type I and Type II kinase inhibitors. We observe that DeepDTAGen performs well on binding predictions for the 16 type II inhibitors in the Davis dataset, but performs poorly on binding predictions for the 34 type II inhibitors against 348 kinases in the Schrodinger dataset.

biophysics↗

bioRxiv2022

Target-template relationships in protein structure prediction and their effect on the accuracy of thermostability calculations

Improving protein thermostability has been a labor- and time-consuming process in industrial applications of protein engineering. Advances in computational approaches have facilitated the development of more efficient strategies to allow the prioritization of stabilizing mutants. Among these is FEP+, a free energy perturbation implementation that uses a thoroughly tested physics-based method to achieve unparalleled accuracy in predicting changes in protein thermostability. To gauge the applicability of FEP+ to situations where crystal structures are unavailable, here we have applied the FEP+ approach to homology models of 12 different proteins covering 316 mutations. By comparing predictions obtained with homology models to those obtained using crystal structures, we have identified that local rather than global sequence conservation between target and template sequence is a determining factor in the accuracy of predictions. By excluding mutation sites with low local sequence identity (<40%) to a template structure, we have obtained predictions with comparable performance to crystal structures (R2 of 0.67 and 0.63 and an RMSE of 1.20 and 1.16 kcal/mol for crystal structure and homology model predictions, respectively) for identifying stabilizing mutations when incorporating residue scanning into a cascade screening strategy. Additionally, we identify and discuss inherent limitations in sequence alignments and homology modeling protocols that translate into the poor FEP+ performance of a few select examples. Overall, our retrospective study provides detailed guidelines for the application of the FEP+ approach using homology models for protein thermostability predictions, which will greatly extend this approach to studies that were previously limited by structure availability.

biophysics↗

bioRxiv2020

Binding Mode of SARS-CoV2 Fusion Peptide to Human Cellular Membrane

Infection of human cells by the SARS-CoV2 relies on its binding to a specific receptor and subsequent fusion of the viral and host cell membranes. The fusion peptide (FP), a short peptide segment in the spike protein, plays a central role in the initial penetration of the virus into the host cell membrane, followed by the fusion of the two membranes. Here, we use an array of molecular dynamics (MD) simulations taking advantage of the Highly Mobile Membrane Mimetic (HMMM) model, to investigate the interaction of the SARS-CoV2 FP with a lipid bilayer representing mammalian cellular membranes at an atomic level, and to characterize the membrane-bound form of the peptide. Six independent systems were generated by changing the initial positioning and orientation of the FP with respect to the membrane, and each system was simulated in five independent replicas, each for 300 ns. In 73% of the simulations, the FP reaches a stable, membrane-bound configuration where the peptide deeply penetrated into the membrane. Clustering of the results reveals three major membrane binding modes (binding modes 1-3) where binding mode 1 populates over half of the data points. Taking into account the sequence conservation among the viral FPs and the results of mutagenesis studies establishing the role of specific residues in the helical portion of the FP in membrane association, the significant depth of penetration of the whole peptide, and the dense population of the respective cluster, we propose that the most deeply inserted membrane-bound form (binding mode 1) represents more closely the biologically relevant form. Analysis of FP-lipid interactions shows the involvement of specific residues, previously described as the "fusion active core residues", in membrane binding. Taken together, the results shed light on a key step involved in SARS-CoV2 infection with potential implications in designing novel inhibitors. SignificanceA key step in cellular infection by the SARS-CoV2 virus is its attachment to and penetration into the plasma membrane of human cells. These processes hinge upon the membrane interaction of the viral fusion peptide, a segment exposed by the spike protein upon its conformational changes after encountering the host cell. In this study, using molecular dynamics simulations, we describe how the fusion peptide from the SARS-CoV2 virus binds human cellular membranes and characterize, at an atomic level, lipid-protein interactions important for the stability of its membrane-bound state.

biophysics↗

bioRxiv2018

Phosphatidic acid inhibits SNARE priming by inducing conformational changes in Sec18 protomers

Eukaryotic homeostasis relies on membrane fusion catalyzed by SNARE proteins. Inactive SNARE bundles are re-activated by Sec18/NSF driven disassembly to enable a new round of fusion. We previously found that phosphatidic acid (PA) binds Sec18 to sequester it from SNAREs. Dephosphorylation of PA dissociates Sec18 from the membrane allowing it to engage SNARE complexes. We now report that PA induces conformational changes in Sec18 protomers, while hexameric Sec18 cannot bind PA membranes. The association of Sec18 with PA was shown to be sensitive to membrane curvature, suggesting that regulation could vary on different organelles in a curvature dependent manner. Molecular dynamics showed that PA binding sites exist on the D1 and D2 domains of Sec18 and that residues needed for binding were masked in the hexameric form of the protein. Together these data indicate that PA regulates Sec18 function through altering protein architecture and stabilizing membrane-bound protomers.

biochemistry↗