AggBERT: Best in Class Prediction of Hexapeptide Amyloidogenesis with a Semi-Supervised ProtBERT Model.

The prediction of peptide amyloidogenesis is a challenging problem in the field of protein folding. Large language models, such as the ProtBERT model, have recently emerged as powerful tools in analyzing protein sequences for applications, such as predicting protein structure and function. In this article, we describe the use of a semisupervised and fine-tuned ProtBERT model to predict peptide amyloidogenesis from sequences alone. Our approach, which we call AggBERT, achieved state-of-the-art performance, demonstrating the potential for large language models to improve the accuracy and speed of amyloid fibril prediction over simple heuristics or structure-based approaches. This work highlights the transformative potential of machine learning and large language models in the fields of chemical biology and biomedicine.

Synergistic mutations in soluble guanylyl cyclase (sGC) reveal a key role for interfacial regions in the sGC activation mechanism.

Soluble guanylyl cyclase (sGC) is the main receptor for nitric oxide (NO) and a central component of the NO-cGMP pathway, critical to cardiovascular function. NO binding to the N-terminal sensor domain in sGC enhances the cyclase activity of the C-terminal catalytic domain. Our understanding of the structural elements regulating this signaling cascade is limited, hindering structure-based drug design efforts that target sGC to improve the management of cardiovascular diseases. Conformational changes are thought to propagate the NO-binding signal throughout the entire sGC heterodimer, via its coiled-coil domain, to reorient the catalytic domain into an active conformation. To identify the structural elements involved in this signal transduction cascade, here we optimized a cGMP-based luciferase assay that reports on heterologous sGC activity in Escherichia coli and identified several mutations that activate sGC. These mutations resided in the dorsal flaps, dimer interface, and GTP-binding regions of the catalytic domain. Combinations of mutations from these different elements synergized, resulting in even greater activity and indicating a complex cross-talk among these regions. Molecular dynamics simulations further revealed conformational changes underlying the functional impact of these mutations. We propose that the interfacial residues play a central role in the sGC activation mechanism by coupling the coiled-coil domain to the active site via a series of hot spots. Our results provide new mechanistic insights not only into the molecular pathway for sGC activation but also for other members of the larger nucleotidyl cyclase family.

A "Clickable" Photoconvertible Small Fluorescent Molecule as a Minimalist Probe for Tracking Individual Biomolecule Complexes.

Photoconvertible fluorophores can enable the visualization and tracking of a specific biomolecules, complexes, and cellular compartments with precise spatiotemporal control. The field of photoconvertible probes is dominated by fluorescent protein variants, which can introduce perturbations to the target biomolecules due to their large size. Here, we present a photoconvertible small molecule, termed CPX, that can be conjugated to any target through azide-alkyne cycloaddition ("click" reaction). To demonstrate its utility, we have applied CPX to study (1) trafficking of biologically relevant synthetic vesicles and (2) intracellular processes involved in transmission of α-synuclein (αS) pathology. Our results demonstrate that CPX can serve as a minimally perturbing probe for tracking the dynamics of biomolecules.

Proteins Need Extra Attention: Improving the Predictive Power of Protein Language Models on Mutational Datasets with Hint Tokens.

In this computational study, we introduce "hint token learning," a novel machine learning approach designed to enhance protein language modeling. This method effectively addresses the unique challenges of protein mutational datasets, characterized by highly similar inputs that may differ by only a single token. Our research highlights the superiority of hint token learning over traditional fine-tuning methods through three distinct case studies. We first developed a highly accurate free energy of folding model using the largest protein stability dataset to date. Then, we applied hint token learning to predict a biophysical attribute, the brightness of green fluorescent protein mutants. In our third case, hint token learning was utilized to assess the impact of mutations on RecA bioactivity. These diverse applications collectively demonstrate the potential of hint token learning for improving protein language modeling across general and specific mutational datasets. To facilitate broader use, we have integrated our protein language models into the HuggingFace ecosystem for downstream, mutational fine-tuning tasks.

Computational Chemistry for the Identification of Lead Compounds for Radiotracer Development.

The use of computer-aided drug design (CADD) for the identification of lead compounds in radiotracer development is steadily increasing. Traditional CADD methods, such as structure-based and ligand-based virtual screening and optimization, have been successfully utilized in many drug discovery programs and are highlighted throughout this review. First, we discuss the use of virtual screening for hit identification at the beginning of drug discovery programs. This is followed by an analysis of how the hits derived from virtual screening can be filtered and culled to highly probable candidates to test in in vitro assays. We then illustrate how CADD can be used to optimize the potency of experimentally validated hit compounds from virtual screening for use in positron emission tomography (PET). Finally, we conclude with a survey of the newest techniques in CADD employing machine learning (ML).

FRETing about the details: Case studies in the use of a genetically encoded fluorescent amino acid for distance-dependent energy transfer.

Förster resonance energy transfer (FRET) is a valuable method for monitoring protein conformation and biomolecular interactions. Intrinsically fluorescent amino acids that can be genetically encoded, such as acridonylalanine (Acd), are particularly useful for FRET studies. However, quantitative interpretation of FRET data to derive distance information requires careful use of controls and consideration of photophysical effects. Here we present two case studies illustrating how Acd can be used in FRET experiments to study small molecule induced conformational changes and multicomponent biomolecular complexes.

Effects of Mutations and Post-Translational Modifications on α-Synuclein In Vitro Aggregation.

Fibrillar aggregates of the α-synuclein (αS) protein are the hallmark of Parkinson's Disease and related neurodegenerative disorders. Characterization of the effects of mutations and post-translational modifications (PTMs) on the αS aggregation rate can provide insight into the mechanism of fibril formation, which remains elusive in spite of intense study. A comprehensive collection (375 examples) of mutant and PTM aggregation rate data measured using the fluorescent probe thioflavin T is presented, as well as a summary of the effects of fluorescent labeling on αS aggregation (20 examples). A curated set of 131 single mutant de novo aggregation experiments are normalized to wild type controls and analyzed in terms of structural data for the monomer and fibrillar forms of αS. These tabulated data serve as a resource to the community to help in interpretation of aggregation experiments and to potentially be used as inputs for computational models of aggregation.

Structural impact of thioamide incorporation into a ß-hairpin.

The thioamide is a naturally-occurring single atom substitution of the canonical amide bond. The exchange of oxygen to sulfur alters the amide's physical and chemical characteristics, thereby expanding its functionality. Incorporation of thioamides in prevalent secondary structures has demonstrated that they can either have stabilizing, destabilizing, or neutral effects. We performed a systematic investigation of the structural impact of thioamide incorporation in a ß-hairpin scaffold with nuclear magnetic resonance (NMR). Thioamides as hydrogen bond donors did not increase the foldedness of the more stable "YKL" variant of this scaffold. In the less stable "HPT" variant of the scaffold, the thioamide could be stabilizing as a hydrogen bond donor and destabilizing as a hydrogen bond acceptor, but the extent of the perturbation depended upon the position of incorporation. To better understand these effects we performed structural modelling of the macrocyclic folded HPT variants. Finally, we compare the thioamide effects that we observe to previous studies of both side-chain and backbone perturbations to this ß-hairpin scaffold to provide context for our observations.

Biomolecular simulation based machine learning models accurately predict sites of tolerability to the unnatural amino acid acridonylalanine.

The incorporation of unnatural amino acids (Uaas) has provided an avenue for novel chemistries to be explored in biological systems. However, the successful application of Uaas is often hampered by site-specific impacts on protein yield and solubility. Although previous efforts to identify features which accurately capture these site-specific effects have been unsuccessful, we have developed a set of novel Rosetta Custom Score Functions and alternative Empirical Score Functions that accurately predict the effects of acridon-2-yl-alanine (Acd) incorporation on protein yield and solubility. Acd-containing mutants were simulated in PyRosetta, and machine learning (ML) was performed using either the decomposed values of the Rosetta energy function, or changes in residue contacts and bioinformatics. Using these feature sets, which represent Rosetta score function specific and bioinformatics-derived terms, ML models were trained to predict highly abstract experimental parameters such as mutant protein yield and solubility and displayed robust performance on well-balanced holdouts. Model feature importance analyses demonstrated that terms corresponding to hydrophobic interactions, desolvation, and amino acid angle preferences played a pivotal role in predicting tolerance of mutation to Acd. Overall, this work provides evidence that the application of ML to features extracted from simulated structural models allow for the accurate prediction of diverse and abstract biological phenomena, beyond the predictivity of traditional modeling and simulation approaches.

Rational design of thioamide peptides as selective inhibitors of cysteine protease cathepsin L.

Aberrant levels of cathepsin L (Cts L), a ubiquitously expressed endosomal cysteine protease, have been implicated in many diseases such as cancer and diabetes. Significantly, Cts L has been identified as a potential target for the treatment of COVID-19 due to its recently unveiled critical role in SARS-CoV-2 entry into the host cells. However, there are currently no clinically approved specific inhibitors of Cts L, as it is often challenging to obtain specificity against the many highly homologous cathepsin family cysteine proteases. Peptide-based agents are often promising protease inhibitors as they offer high selectivity and potency, but unfortunately are subject to degradation in vivo. Thioamide substitution, a single-atom O-to-S modification in the peptide backbone, has been shown to improve the proteolytic stability of peptides addressing this issue. Utilizing this approach, we demonstrate herein that good peptidyl substrates can be converted into sub-micromolar inhibitors of Cts L by a single thioamide substitution in the peptide backbone. We have designed and scanned several thioamide stabilized peptide scaffolds, in which one peptide, RS 1A, was stabilized against proteolysis by all five cathepsins (Cts L, Cts V, Cts K, Cts S, and Cts B) while inhibiting Cts L with >25-fold specificity against the other cathepsins. We further showed that this stabilized RS 1A peptide could inhibit Cts L in human liver carcinoma lysates (IC50 = 19 µM). Our study demonstrates that one can rationally design a stabilized, specific peptidyl protease inhibitor by strategic placement of a thioamide and reaffirms the place of this single-atom modification in the toolbox of peptide-based rational drug design.

Correction: Rosetta custom score functions accurately predict ΔΔG of mutations at protein-protein interfaces using machine learning.

Correction for 'Rosetta custom score functions accurately predict ΔΔG of mutations at protein-protein interfaces using machine learning' by Sumant R. Shringari et al., Chem. Commun., 2020, 56, 6774-6777, DOI: .

Rosetta custom score functions accurately predict ΔΔG of mutations at protein-protein interfaces using machine learning.

Protein-protein interfaces play essential roles in a variety of biological processes and many therapeutic molecules are targeted at these interfaces. However, accurate predictions of the effects of interfacial mutations to identify "hotspots" have remained elusive despite the myriad of modeling and machine learning methods tested. Here, for the first time, we demonstrate that nonlinear reweighting of energy terms from Rosetta, through the use of machine learning, exhibits improved predictability of ΔΔG values associated with interfacial mutations.

Rosetta Machine Learning Models Accurately Classify Positional Effects of Thioamides on Proteolysis.

Thioamide substitutions of the peptide backbone have been shown to stabilize therapeutic and imaging peptides toward proteolysis. In order to rationally design thioamide modifications, we have developed a novel Rosetta custom score function to classify thioamide positional effects on proteolysis in substrates of serine and cysteine proteases. Peptides of interest were docked into proteases using the FlexPepDock application in Rosetta. Docked complexes were modified to contain thioamides parametrized through the creation of custom atom types in Rosetta based on ab intio simulations. Thioamide complexes were simulated, and the resultant structural complexes provided features for machine learning classification as the decomposed values of the Rosetta score function. An ensemble, majority voting model was developed to be a robust predictor of previously unpublished thioamide proteolysis holdout data. Theoretical control simulations with pseudo-atoms that modulate only one physical characteristic of the thioamide show differential effects on prediction accuracy by the optimized voting classification model. These pseudo-atom model simulations, as well as statistical analyses of the full thioamide simulations, implicate steric effects on peptide binding as being primarily responsible for thioamide positional effects on proteolytic resistance.

Studies of Thioamide Effects on Serine Protease Activity Enable Two-Site Stabilization of Cancer Imaging Peptides.

Thioamide substitutions in peptides can be used as fluorescence quenchers in protease sensors and as stabilizing modifications of hormone analogs. To guide these applications in the context of serine proteases, we here examine the cleavage of several model substrates, scanning a thioamide between the P3 and P3' positions, and identify perturbing positions for thioamide substitution. While all serine proteases tested were affected by P1 thioamidation, certain proteases were also significantly affected by other thioamide positions. We demonstrate how these findings can be applied by harnessing the combined P3/P1 effect of a single thioamide on kallikrein proteolysis to protect two key positions in a neuropeptide Y-based imaging probe, increasing its serum half-life to >24 h while maintaining potency for binding to Y1 receptor expressing cells. Such stabilized peptide probes could find application in imaging cell populations in animal models or even in clinical applications such as fluorescence-guided surgery.

Identification of a nanomolar affinity α-synuclein fibril imaging probe by ultra-high throughput in silico screening.

Small molecules that bind with high affinity and specificity to fibrils of the α-synuclein (αS) protein have the potential to serve as positron emission tomography (PET) imaging probes to aid in the diagnosis of Parkinson's disease and related synucleinopathies. To identify such molecules, we employed an ultra-high throughput in silico screening strategy using idealized pseudo-ligands termed exemplars to identify compounds for experimental binding studies. For the top hit from this screen, we used photo-crosslinking to confirm its binding site and studied the structure-activity relationship of its analogs to develop multiple molecules with nanomolar affinity for αS fibrils and moderate specificity for αS over Aß fibrils. Lastly, we demonstrated the potential of the lead analog as an imaging probe by measuring binding to αS-enriched homogenates from mouse brain tissue using a radiolabeled analog of the identified molecule. This study demonstrates the validity of our powerful new approach to the discovery of PET probes for challenging molecular targets.