Taras V. Pogorelov

The Wnt protein family plays a critical role in cell development, with each Wnt protein interacting differently with the Wntless (Wls) membrane protein through distinct binding residues. A direct comparison and elucidation of the molecular mechanisms underlying Wnt–Wls binding across the diverse Wnt family remain challenging, owing to variations in sequence length and amino acid composition among Wnt proteins, which can affect their binding affinity and trafficking efficiency via Wls. Here we combine atomistic molecular dynamics simulations with supervised machine learning to elucidate binding specificity among four Wnt proteins, selected based on experimental structure availability and scientific relevance. We implement a local structure alignment algorithm to enable cross-system matching and comparison of residue interactions, and we apply a two-stage clustering strategy to reduce feature redundancy and facilitate robust feature selection. After training a Random Forest classifier that achieved high predicting accuracy, our feature importance analysis reveals both previously known and novel key residue pairs responsible for distinguishing among the Wnt systems. Our findings highlight that the binding specificity across different systems arises from the distributed nature of interactions across the protein binding surface and demonstrate how interpretable machine learning can effectively uncover crucial biophysical interactions. Importantly, our integrated strategy is generalizable to other systems and provides a data-driven approach for analyzing protein–protein interactions and guiding experimental validation or therapeutic targeting.

The ion channel-forming natural product amphotericin B (AmB) can serve as a molecular prosthetic for the cystic fibrosis transmembrane conductance regulator (CFTR) anion channel and thereby restore host defenses in cultured cystic fibrosis (CF) airway epithelia. This is despite the fact that the permeability of AmB-based channels favors cations, and these channels lose their capacity to increase airway surface liquid (ASL) pH in CF airway epithelia at high concentrations. We hypothesize that modifying such channels to favor anion permeability would make them more CFTR-like and thus increase their potential therapeutic effects compared to AmB. Here we show that a synthetic derivative of AmB, AmB-AA, which has an added positively charged appendage and forms ion channels with an improved relative permeability to anions, outperformed AmB in increasing the ASL pH in CF airway epithelia at both low and high concentrations. Further modifications led to another AmB derivative, C2’epiAmB-AA, that also minimized cholesterol binding and thus toxicity to cultured CF airway epithelia and was an effective surrogate for CFTR in primary cultured airway epithelia from people with CF.

The Wnt protein family plays a critical role in cell development, with each Wnt protein interacting differently with the Wls membrane protein through distinct binding residues. A direct comparison and elucidation of the molecular mechanisms underlying Wnt–Wls binding across the diverse Wnt family remain challenging, owing to variations in sequence length and amino acid composition among Wnt proteins, which can affect their binding affinity and trafficking efficiency via Wls. Here we combine extended atomistic molecular dynamics simulations with supervised machine learning to elucidate binding specificity among four Wnt proteins, selected based on experimental structure availability and scientific relevance. We implement a local structure alignment algorithm to enable cross-system matching and comparison of residue interactions, and we apply a two-stage clustering strategy to reduce feature redundancy and facilitate robust feature selection. After training a Random Forest classifier that achieved high predicting accuracy, our feature importance analysis reveals both previously known and novel key residue pairs responsible for distinguishing among the Wnt systems. Our findings highlight that the binding specificity across different systems arises from the distributed nature of interactions across the protein binding surface and demonstrate how interpretable machine learning can effectively uncover crucial biophysical interactions. Importantly, our integrated strategy is generalizable to other systems and provides a data-driven approach for analyzing protein– protein interactions and guiding experimental validation or therapeutic targeting.

Antimicrobial peptides (AMPs) have emerged as a promising solution to the escalating public health threat caused by multidrug-resistant bacteria. Although ongoing research efforts have established AMP's role in membrane permeabilization and leakage, the precise mechanisms driving these disruption patterns remain unclear. We leverage molecular dynamics (MD) simulations enhanced by membrane mimetic (HMMM) to systematically investigate how the physiochemical properties of magainin (+3) and pexiganan (+9) affect their localization, insertion, curvature perturbation, and membrane binding ensemble. Building on existing microbiology, NMR, circular dichroism, and fluorescence data, our analysis reveals that the lipid makeup is a key determinant in the binding dynamics and structural conformation of AMPs. We find that phospholipid type is crucial for peptide localization, demonstrated through magainin's predominant interaction with lipid tails and pexiganan's with polar headgroups in POPC/POPS membranes. The membrane curvature changes induced by pexiganan relative to magainin suggest that AMPs with larger charges have more potential in modulating bilayer bending. These insights advance our understanding of AMP-membrane interactions at the molecular level, offering guidance for the design of targeted antimicrobial therapies.

Enzyme–enzyme interactions are fundamental to the function of cells. Their atomistic mechanisms remain elusive mainly due to limitations of in-cell measurements. We address this challenge by atomistically modeling, for a total of ≈80 μs, a slice of the human cell cytoplasm that includes three successive enzymes along the glycolytic pathway: glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphoglycerate kinase (PGK), and phosphoglycerate mutase (PGM). We tested the model for nonspecific protein stickiness, an artifact of current atomistic force fields in crowded environments. The simulations reveal that the human enzymes co-organize in-cell into transient submetabolon complexes, consistent with previous experimental results. Our data both reiterate known specificity between GAPDH and PGK and reveal extensive direct interactions between GAPDH and PGM. Our simulations further reveal, through force field benchmarking, the critical role of protein solvation in facilitating these enzyme–enzyme interactions. Transient interenzyme interactions with μs lifetime occur repeatedly in our simulations via specific sticky protein surface patches, with interactions often mediated by charged patch residues. Some of the residues that interact frequently with one another lie in or near the active site of the enzymes. We show that some of these patches correspond to a general mode to interact with several partners for promiscuous enzymes like GAPDH. We further show that the non-native yeast PGK is stickier than human PGK in our human cytoplasm model, supporting the idea of evolutionary pressure to reduce sticking. Our cytoplasm modeling paves the way toward capturing the atomistic dynamics of an entire enzymatic pathway in-cell.

Methane emissions from livestock contribute to global warming. Seaweeds used as food additive offer a promising emission mitigation strategy because seaweeds are enriched in bromoform─a methanogenesis inhibitor. Therefore, understanding bromoform storage and production in seaweeds and particularly in a cell-like environment is crucial. As a first step toward this aim, we present an atomistic description of bromoform dynamics, diffusion, and aggregation in the presence of lipid membranes. Using all-atom molecular dynamics simulations with customized CHARMM-formatted bromoform force field files, we investigate the interactions of bromoform and lipid bilayer across various concentrations. Bromoform penetrates membranes and at high concentrations forms aggregates outside the membrane without affecting membrane thickness or lipid tail order. Aggregates outside the membrane influence the membrane curvature. Within the membrane, bromoform preferentially localizes in the membrane hydrophobic core and diffuses the slowest along the membrane normal. Employing general local-atomic descriptors and unsupervised machine learning, we demonstrate the similarity of bromoform local structures between the liquid and aggregated forms.

Protein–protein and protein–water hydrogen bonding interactions play essential roles in the way a protein passes through the transition state during folding or unfolding, but the large number of these interactions in molecular dynamics (MD) simulations makes them difficult to analyze. Here, we introduce a state space representation and associated “rarity” measure to identify and quantify transition state passage (transit) events. Applying this representation to a long MD simulation trajectory that captured multiple folding and unfolding events of the GTT WW domain, a small protein often used as a model for the folding process, we identified three transition categories: Highway (faster), Meander (slower), and Ambiguous (intermediate). We developed data sonification and visualization tools to analyze hydrogen bond dynamics before, during, and after these transition events. By means of these tools, we were able to identify characteristic hydrogen bonding patterns associated with “Highway” versus “Meander” versus “Ambiguous” transitions and to design algorithms that can identify these same folding pathways and critical protein–water interactions directly from the data. Highly cooperative hydrogen bonding can either slow down or speed up transit. Furthermore, an analysis of protein–water hydrogen bond dynamics at the surface of WW domain shows an increase in hydrogen bond lifetime from folded to unfolded conformations with Ambiguous transitions as an outlier. In summary, hydrogen bond dynamics provide a direct window into the heterogeneity of transits, which can vary widely in duration (by a factor of 10) due to a complex energy landscape.

Many enzymes undergo major conformational changes to function in cells, particularly when they bind to more than one substrate. We quantify the large-amplitude hinge-bending landscape of human phosphoglycerate kinase (PGK) in a human cytoplasm. Approximately 70 μs of all-atom simulations, upon coarse graining, reveal three metastable states of PGK with different hinge angle distributions and additional substates. The "open" state was more populated than the "semi-open" or "closed" states. In addition to free energies and barriers within the landscape, we characterized the average transition state passage time of ≈0.3 μs and reversible substrate and product binding. Human PGK in a dilute solution simulation shows a transition directly from the open to closed states, in agreement with previous SAXS experiments, suggesting that the cell-like model environment promotes stability of the human PGK semi-open state. Yeast PGK also sampled three metastable states within the cytoplasm model, with the closed state favored in our simulation.

Decades of previous efforts to develop renal-sparing polyene antifungals were misguided by the classic membrane permeabilization model¹. Recently, the clinically vital but also highly renal-toxic small-molecule natural product amphotericin B was instead found to kill fungi primarily by forming extramembraneous sponge-like aggregates that extract ergosterol from lipid bilayers^2,3,4,5,6. Here we show that rapid and selective extraction of fungal ergosterol can yield potent and renal-sparing polyene antifungals. Cholesterol extraction was found to drive the toxicity of amphotericin B to human renal cells. Our examination of high-resolution structures of amphotericin B sponges in sterol-free and sterol-bound states guided us to a promising structural derivative that does not bind cholesterol and is thus renal sparing. This derivative was also less potent because it extracts ergosterol more slowly. Selective acceleration of ergosterol extraction with a second structural modification yielded a new polyene, AM-2-19, that is renal sparing in mice and primary human renal cells, potent against hundreds of pathogenic fungal strains, resistance evasive following serial passage in vitro and highly efficacious in animal models of invasive fungal infections. Thus, rational tuning of the dynamics of interactions between small molecules may lead to better treatments for fungal infections that still kill millions of people annually^7,8 and potentially other resistance-evasive antimicrobials, including those that have recently been shown to operate through supramolecular structures that target specific lipids⁹.