DDMut-PPI: predicting effects of mutations on protein-protein interactions using graph-based deep learning.

Protein-protein interactions (PPIs) play a vital role in cellular functions and are essential for therapeutic development and understanding diseases. However, current predictive tools often struggle to balance efficiency and precision in predicting the effects of mutations on these complex interactions. To address this, we present DDMut-PPI, a deep learning model that efficiently and accurately predicts changes in PPI binding free energy upon single and multiple point mutations. Building on the robust Siamese network architecture with graph-based signatures from our prior work, DDMut, the DDMut-PPI model was enhanced with a graph convolutional network operated on the protein interaction interface. We used residue-specific embeddings from ProtT5 protein language model as node features, and a variety of molecular interactions as edge features. By integrating evolutionary context with spatial information, this framework enables DDMut-PPI to achieve a robust Pearson correlation of up to 0.75 (root mean squared error: 1.33 kcal/mol) in our evaluations, outperforming most existing methods. Importantly, the model demonstrated consistent performance across mutations that increase or decrease binding affinity. DDMut-PPI offers a significant advancement in the field and will serve as a valuable tool for researchers probing the complexities of protein interactions. DDMut-PPI is freely available as a web server and an application programming interface at https://biosig.lab.uq.edu.au/ddmut_ppi.

DDMut: predicting effects of mutations on protein stability using deep learning.

Understanding the effects of mutations on protein stability is crucial for variant interpretation and prioritisation, protein engineering, and biotechnology. Despite significant efforts, community assessments of predictive tools have highlighted ongoing limitations, including computational time, low predictive power, and biased predictions towards destabilising mutations. To fill this gap, we developed DDMut, a fast and accurate siamese network to predict changes in Gibbs Free Energy upon single and multiple point mutations, leveraging both forward and hypothetical reverse mutations to account for model anti-symmetry. Deep learning models were built by integrating graph-based representations of the localised 3D environment, with convolutional layers and transformer encoders. This combination better captured the distance patterns between atoms by extracting both short-range and long-range interactions. DDMut achieved Pearson's correlations of up to 0.70 (RMSE: 1.37 kcal/mol) on single point mutations, and 0.70 (RMSE: 1.84 kcal/mol) on double/triple mutants, outperforming most available methods across non-redundant blind test sets. Importantly, DDMut was highly scalable and demonstrated anti-symmetric performance on both destabilising and stabilising mutations. We believe DDMut will be a useful platform to better understand the functional consequences of mutations, and guide rational protein engineering. DDMut is freely available as a web server and API at https://biosig.lab.uq.edu.au/ddmut.

Exploring the effects of missense mutations on protein thermodynamics through structure-based approaches: findings from the CAGI6 challenges.

Missense mutations are known contributors to diverse genetic disorders, due to their subtle, single amino acid changes imparted on the resultant protein. Because of this, understanding the impact of these mutations on protein stability and function is crucial for unravelling disease mechanisms and developing targeted therapies. The Critical Assessment of Genome Interpretation (CAGI) provides a valuable platform for benchmarking state-of-the-art computational methods in predicting the impact of disease-related mutations on protein thermodynamics. Here we report the performance of our comprehensive platform of structure-based computational approaches to evaluate mutations impacting protein structure and function on 3 challenges from CAGI6: Calmodulin, MAPK1 and MAPK3. Our stability predictors have achieved correlations of up to 0.74 and AUCs of 1 when predicting changes in ΔΔG for MAPK1 and MAPK3, respectively, and AUC of up to 0.75 in the Calmodulin challenge. Overall, our study highlights the importance of structure-based approaches in understanding the effects of missense mutations on protein thermodynamics. The results obtained from the CAGI6 challenges contribute to the ongoing efforts to enhance our understanding of disease mechanisms and facilitate the development of personalised medicine approaches.

Structural landscapes of PPI interfaces.

Proteins are capable of highly specific interactions and are responsible for a wide range of functions, making them attractive in the pursuit of new therapeutic options. Previous studies focusing on overall geometry of protein-protein interfaces, however, concluded that PPI interfaces were generally flat. More recently, this idea has been challenged by their structural and thermodynamic characterisation, suggesting the existence of concave binding sites that are closer in character to traditional small-molecule binding sites, rather than exhibiting complete flatness. Here, we present a large-scale analysis of binding geometry and physicochemical properties of all protein-protein interfaces available in the Protein Data Bank. In this review, we provide a comprehensive overview of the protein-protein interface landscape, including evidence that even for overall larger, more flat interfaces that utilize discontinuous interacting regions, small and potentially druggable pockets are utilized at binding sites.

DockNet: high-throughput protein-protein interface contact prediction.

MOTIVATION: Over 300 000 protein-protein interaction (PPI) pairs have been identified in the human proteome and targeting these is fast becoming the next frontier in drug design. Predicting PPI sites, however, is a challenging task that traditionally requires computationally expensive and time-consuming docking simulations. A major weakness of modern protein docking algorithms is the inability to account for protein flexibility, which ultimately leads to relatively poor results. RESULTS: Here, we propose DockNet, an efficient Siamese graph-based neural network method which predicts contact residues between two interacting proteins. Unlike other methods that only utilize a protein's surface or treat the protein structure as a rigid body, DockNet incorporates the entire protein structure and places no limits on protein flexibility during an interaction. Predictions are modeled at the residue level, based on a diverse set of input node features including residue type, surface accessibility, residue depth, secondary structure, pharmacophore and torsional angles. DockNet is comparable to current state-of-the-art methods, achieving an area under the curve (AUC) value of up to 0.84 on an independent test set (DB5), can be applied to a variety of different protein structures and can be utilized in situations where accurate unbound protein structures cannot be obtained. AVAILABILITY AND IMPLEMENTATION: DockNet is available at https://github.com/npwilliams09/docknet and an easy-to-use webserver at https://biosig.lab.uq.edu.au/docknet. All other data underlying this article are available in the article and in its online supplementary material. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

LEGO-CSM: a tool for functional characterization of proteins.

MOTIVATION: With the development of sequencing techniques, the discovery of new proteins significantly exceeds the human capacity and resources for experimentally characterizing protein functions. Localization, EC numbers, and GO terms with the structure-based Cutoff Scanning Matrix (LEGO-CSM) is a comprehensive web-based resource that fills this gap by leveraging the well-established and robust graph-based signatures to supervised learning models using both protein sequence and structure information to accurately model protein function in terms of Subcellular Localization, Enzyme Commission (EC) numbers, and Gene Ontology (GO) terms. RESULTS: We show our models perform as well as or better than alternative approaches, achieving area under the receiver operating characteristic curve of up to 0.93 for subcellular localization, up to 0.93 for EC, and up to 0.81 for GO terms on independent blind tests. AVAILABILITY AND IMPLEMENTATION: LEGO-CSM's web server is freely available at https://biosig.lab.uq.edu.au/lego_csm. In addition, all datasets used to train and test LEGO-CSM's models can be downloaded at https://biosig.lab.uq.edu.au/lego_csm/data.

CSM-Potential: mapping protein interactions and biological ligands in 3D space using geometric deep learning.

Recent advances in protein structural modelling have enabled the accurate prediction of the holo 3D structures of almost any protein, however protein function is intrinsically linked to the interactions it makes. While a number of computational approaches have been proposed to explore potential biological interactions, they have been limited to specific interactions, and have not been readily accessible for non-experts or use in bioinformatics pipelines. Here we present CSM-Potential, a geometric deep learning approach to identify regions of a protein surface that are likely to mediate protein-protein and protein-ligand interactions in order to provide a link between 3D structure and biological function. Our method has shown robust performance, outperforming existing methods for both predictive tasks. By assessing the performance of CSM-Potential on independent blind tests, we show that our method was able to achieve ROC AUC values of up to 0.81 for the identification of potential protein-protein binding sites, and up to 0.96 accuracy on biological ligand classification. Our method is freely available as a user-friendly and easy-to-use web server and API at http://biosig.unimelb.edu.au/csm_potential.

CSM-Potential2: A comprehensive deep learning platform for the analysis of protein interacting interfaces.

Proteins are molecular machinery that participate in virtually all essential biological functions within the cell, which are tightly related to their 3D structure. The importance of understanding protein structure-function relationship is highlighted by the exponential growth of experimental structures, which has been greatly expanded by recent breakthroughs in protein structure prediction, most notably RosettaFold, and AlphaFold2. These advances have prompted the development of several computational approaches that leverage these data sources to explore potential biological interactions. However, most methods are generally limited to analysis of single types of interactions, such as protein-protein or protein-ligand interactions, and their complexity limits the usability to expert users. Here we report CSM-Potential2, a deep learning platform for the analysis of binding interfaces on protein structures. In addition to prediction of protein-protein interactions binding sites and classification of biological ligands, our new platform incorporates prediction of interactions with nucleic acids at the residue level and allows for ligand transplantation based on sequence and structure similarity to experimentally determined structures. We anticipate our platform to be a valuable resource that provides easy access to a range of state-of-the-art methods to expert and non-expert users for the study of biological interactions. Our tool is freely available as an easy-to-use web server and API available at https://biosig.lab.uq.edu.au/csm_potential.

mmCSM-PPI: predicting the effects of multiple point mutations on protein-protein interactions.

Protein-protein interactions play a crucial role in all cellular functions and biological processes and mutations leading to their disruption are enriched in many diseases. While a number of computational methods to assess the effects of variants on protein-protein binding affinity have been proposed, they are in general limited to the analysis of single point mutations and have been shown to perform poorly on independent test sets. Here, we present mmCSM-PPI, a scalable and effective machine learning model for accurately assessing changes in protein-protein binding affinity caused by single and multiple missense mutations. We expanded our well-established graph-based signatures in order to capture physicochemical and geometrical properties of multiple wild-type residue environments and integrated them with substitution scores and dynamics terms from normal mode analysis. mmCSM-PPI was able to achieve a Pearson's correlation of up to 0.75 (RMSE = 1.64 kcal/mol) under 10-fold cross-validation and 0.70 (RMSE = 2.06 kcal/mol) on a non-redundant blind test, outperforming existing methods. Our method is freely available as a user-friendly and easy-to-use web server and API at http://biosig.unimelb.edu.au/mmcsm_ppi.

MTR3D: identifying regions within protein tertiary structures under purifying selection.

The identification of disease-causal variants is non-trivial. By mapping population variation from over 448,000 exome and genome sequences to over 81,000 experimental structures and homology models of the human proteome, we have calculated both regional intolerance to missense variation (Missense Tolerance Ratio, MTR), using a sliding window of 21-41 codons, and introduce a new 3D spatial intolerance to missense variation score (3D Missense Tolerance Ratio, MTR3D), using spheres of 5-8 Å. We show that the MTR3D is less biased by regions with limited data and more accurately identifies regions under purifying selection than estimates relying on the sequence alone. Intolerant regions were highly enriched for both ClinVar pathogenic and COSMIC somatic missense variants (Mann-Whitney U test P < 2.2 × 10-16). Further, we combine sequence- and spatial-based scores to generate a consensus score, MTRX, which distinguishes pathogenic from benign variants more accurately than either score separately (AUC = 0.85). The MTR3D server enables easy visualisation of population variation, MTR, MTR3D and MTRX scores across the entire gene and protein structure for >17,000 human genes and >42,000 alternative alternate transcripts, including both Ensembl and RefSeq transcripts. MTR3D is freely available by user-friendly web-interface and API at http://biosig.unimelb.edu.au/mtr3d/.

mCSM-membrane: predicting the effects of mutations on transmembrane proteins.

Significant efforts have been invested into understanding and predicting the molecular consequences of mutations in protein coding regions, however nearly all approaches have been developed using globular, soluble proteins. These methods have been shown to poorly translate to studying the effects of mutations in membrane proteins. To fill this gap, here we report, mCSM-membrane, a user-friendly web server that can be used to analyse the impacts of mutations on membrane protein stability and the likelihood of them being disease associated. mCSM-membrane derives from our well-established mutation modelling approach that uses graph-based signatures to model protein geometry and physicochemical properties for supervised learning. Our stability predictor achieved correlations of up to 0.72 and 0.67 (on cross validation and blind tests, respectively), while our pathogenicity predictor achieved a Matthew's Correlation Coefficient (MCC) of up to 0.77 and 0.73, outperforming previously described methods in both predicting changes in stability and in identifying pathogenic variants. mCSM-membrane will be an invaluable and dedicated resource for investigating the effects of single-point mutations on membrane proteins through a freely available, user friendly web server at http://biosig.unimelb.edu.au/mcsm_membrane.

mCSM-AB2: guiding rational antibody design using graph-based signatures.

MOTIVATION: A lack of accurate computational tools to guide rational mutagenesis has made affinity maturation a recurrent challenge in antibody (Ab) development. We previously showed that graph-based signatures can be used to predict the effects of mutations on Ab binding affinity. RESULTS: Here we present an updated and refined version of this approach, mCSM-AB2, capable of accurately modelling the effects of mutations on Ab-antigen binding affinity, through the inclusion of evolutionary and energetic terms. Using a new and expanded database of over 1800 mutations with experimental binding measurements and structural information, mCSM-AB2 achieved a Pearson's correlation of 0.73 and 0.77 across training and blind tests, respectively, outperforming available methods currently used for rational Ab engineering. AVAILABILITY AND IMPLEMENTATION: mCSM-AB2 is available as a user-friendly and freely accessible web server providing rapid analysis of both individual mutations or the entire binding interface to guide rational antibody affinity maturation at http://biosig.unimelb.edu.au/mcsm_ab2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

EasyVS: a user-friendly web-based tool for molecule library selection and structure-based virtual screening.

SUMMARY: EasyVS is a web-based platform built to simplify molecule library selection and virtual screening. With an intuitive interface, the tool allows users to go from selecting a protein target with a known structure and tailoring a purchasable molecule library to performing and visualizing docking in a few clicks. Our system also allows users to filter screening libraries based on molecule properties, cluster molecules by similarity and personalize docking parameters. AVAILABILITY AND IMPLEMENTATION: EasyVS is freely available as an easy-to-use web interface at http://biosig.unimelb.edu.au/easyvs. CONTACT: douglas.pires@unimelb.edu.au or david.ascher@unimelb.edu.au. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

pdCSM-PPI: Using Graph-Based Signatures to Identify Protein-Protein Interaction Inhibitors.

Protein-protein interactions are promising sites for development of selective drugs; however, they have generally been viewed as challenging targets. Molecules targeting protein-protein interactions tend to be larger and more lipophilic than other drug-like molecules, mimicking the properties of interacting interfaces. Here, we propose a machine learning approach that uses a graph-based representation of small molecules to guide identification of inhibitors modulating protein-protein interactions, pdCSM-PPI. This approach was applied to 21 different PPI targets. We developed interaction-specific models that were able to accurately identify active compounds achieving MCC and F1 scores up to 1, and Pearson's correlations up to 0.87, outperforming previous approaches. Using insights from these individual models, we developed a generic protein-protein interaction modulator predictive model, which accurately predicted IC50 with a Pearson's correlation of 0.64 on a low redundancy blind test. Importantly, we were able to accurately identify active from inactive compounds, achieving an AUC of 0.77 and sensitivity and specificity of 76% and 78%, respectively. We believe pdCSM-PPI will be an important tool to help guide more efficient screening of new PPI inhibitors; it is freely available as an easy-to-use web server and API at http://biosig.unimelb.edu.au/pdcsm_ppi.

mCSM-PPI2: predicting the effects of mutations on protein-protein interactions.

Protein-protein Interactions are involved in most fundamental biological processes, with disease causing mutations enriched at their interfaces. Here we present mCSM-PPI2, a novel machine learning computational tool designed to more accurately predict the effects of missense mutations on protein-protein interaction binding affinity. mCSM-PPI2 uses graph-based structural signatures to model effects of variations on the inter-residue interaction network, evolutionary information, complex network metrics and energetic terms to generate an optimised predictor. We demonstrate that our method outperforms previous methods, ranking first among 26 others on CAPRI blind tests. mCSM-PPI2 is freely available as a user friendly webserver at http://biosig.unimelb.edu.au/mcsm_ppi2/.

EFG-CS: Predicting chemical shifts from amino acid sequences with protein structure prediction using machine learning and deep learning models.

Nuclear magnetic resonance (NMR) crystallography is one of the main methods in structural biology for analyzing protein stereochemistry and structure. The chemical shift of the resonance frequency reflects the effect of the protons in a molecule producing distinct NMR signals in different chemical environments. Apprehending chemical shifts from NMR signals can be challenging since having an NMR structure does not necessarily provide all the required chemical shift information, making predictive models essential for accurately deducing chemical shifts, either from protein structures or, more ideally, directly from amino acid sequences. Here, we present EFG-CS, a web server that specializes in chemical shift prediction. EFG-CS employs a machine learning-based transfer prediction model for backbone atom chemical shift prediction, using ESMFold-predicted protein structures. Additionally, ESG-CS incorporates a graph neural network-based model to provide comprehensive side-chain atom chemical shift predictions. Our method demonstrated reliable performance in backbone atom prediction, achieving comparable accuracy levels with root mean square errors (RMSE) of 0.30 ppm for H, 0.22 ppm for Hα, 0.89 ppm for C, 0.89 ppm for Cα, 0.84 ppm for Cß, and 1.69 ppm for N. Moreover, our approach also showed predictive capabilities in side-chain atom chemical shift prediction achieving RMSE values of 0.71 ppm for Hß, 0.74-1.15 ppm for Hδ, and 0.58-0.94 ppm for Hγ, solely utilizing amino acid sequences without homology or feature curation. This work shows for the first time that generative AI protein models can predict NMR shifts nearly comparable to experimental models. This web server is freely available at https://biosig.lab.uq.edu.au/efg_cs, and the chemical shift prediction results can be downloaded in tabular format and visualized in 3D format.

MTR3D-AF2: Expanding the coverage of spatially derived missense tolerance scores across the human proteome using AlphaFold2.

The missense tolerance ratio (MTR) was developed as a novel approach to assess the deleteriousness of variants. Its three-dimensional successor, MTR3D, was demonstrated powerful at discriminating pathogenic from benign variants. However, its reliance on experimental structures and homologs limited its coverage of the proteome. We have now utilized AlphaFold2 models to develop MTR3D-AF2, which covers 89.31% of proteins and 85.39% of residues across the human proteome. This work has improved MTR3D's ability to distinguish clinically established pathogenic from benign variants. MTR3D-AF2 is freely available as an interactive web server at https://biosig.lab.uq.edu.au/mtr3daf2/.

piscesCSM: prediction of anticancer synergistic drug combinations.

While drug combination therapies are of great importance, particularly in cancer treatment, identifying novel synergistic drug combinations has been a challenging venture. Computational methods have emerged in this context as a promising tool for prioritizing drug combinations for further evaluation, though they have presented limited performance, utility, and interpretability. Here, we propose a novel predictive tool, piscesCSM, that leverages graph-based representations to model small molecule chemical structures to accurately predict drug combinations with favourable anticancer synergistic effects against one or multiple cancer cell lines. Leveraging these insights, we developed a general supervised machine learning model to guide the prediction of anticancer synergistic drug combinations in over 30 cell lines. It achieved an area under the receiver operating characteristic curve (AUROC) of up to 0.89 on independent non-redundant blind tests, outperforming state-of-the-art approaches on both large-scale oncology screening data and an independent test set generated by AstraZeneca (with more than a 16% improvement in predictive accuracy). Moreover, by exploring the interpretability of our approach, we found that simple physicochemical properties and graph-based signatures are predictive of chemotherapy synergism. To provide a simple and integrated platform to rapidly screen potential candidate pairs with favourable synergistic anticancer effects, we made piscesCSM freely available online at https://biosig.lab.uq.edu.au/piscescsm/ as a web server and API. We believe that our predictive tool will provide a valuable resource for optimizing and augmenting combinatorial screening libraries to identify effective and safe synergistic anticancer drug combinations. SCIENTIFIC CONTRIBUTION: This work proposes piscesCSM, a machine-learning-based framework that relies on well-established graph-based representations of small molecules to identify and provide better predictive accuracy of syngenetic drug combinations. Our model, piscesCSM, shows that combining physiochemical properties with graph-based signatures can outperform current architectures on classification prediction tasks. Furthermore, implementing our tool as a web server offers a user-friendly platform for researchers to screen for potential synergistic drug combinations with favorable anticancer effects against one or multiple cancer cell lines.

CSM-Toxin: A Web-Server for Predicting Protein Toxicity.

Biologics are one of the most rapidly expanding classes of therapeutics, but can be associated with a range of toxic properties. In small-molecule drug development, early identification of potential toxicity led to a significant reduction in clinical trial failures, however we currently lack robust qualitative rules or predictive tools for peptide- and protein-based biologics. To address this, we have manually curated the largest set of high-quality experimental data on peptide and protein toxicities, and developed CSM-Toxin, a novel in-silico protein toxicity classifier, which relies solely on the protein primary sequence. Our approach encodes the protein sequence information using a deep learning natural languages model to understand "biological" language, where residues are treated as words and protein sequences as sentences. The CSM-Toxin was able to accurately identify peptides and proteins with potential toxicity, achieving an MCC of up to 0.66 across both cross-validation and multiple non-redundant blind tests, outperforming other methods and highlighting the robust and generalisable performance of our model. We strongly believe the CSM-Toxin will serve as a valuable platform to minimise potential toxicity in the biologic development pipeline. Our method is freely available as an easy-to-use webserver.

HGDiscovery: An online tool providing functional and phenotypic information on novel variants of homogentisate 1,2- dioxigenase.

Alkaptonuria (AKU), a rare genetic disorder, is characterized by the accumulation of homogentisic acid (HGA) in the body. Affected individuals lack functional levels of an enzyme required to breakdown HGA. Mutations in the homogentisate 1,2-dioxygenase (HGD) gene cause AKU and they are responsible for deficient levels of functional HGD, which, in turn, leads to excess levels of HGA. Although HGA is rapidly cleared from the body by the kidneys, in the long term it starts accumulating in various tissues, especially cartilage. Over time (rarely before adulthood), it eventually changes the color of affected tissue to slate blue or black. Here we report a comprehensive mutation analysis of 111 pathogenic and 190 non-pathogenic HGD missense mutations using protein structural information. Using our comprehensive suite of graph-based signature methods, mCSM complemented with sequence-based tools, we studied the functional and molecular consequences of each mutation on protein stability, interaction and evolutionary conservation. The scores generated from the structure and sequence-based tools were used to train a supervised machine learning algorithm with 89% accuracy. The empirical classifier was used to generate the variant phenotype for novel HGD missense mutations. All this information is deployed as a user friendly freely available web server called HGDiscovery (https://biosig.lab.uq.edu.au/hgdiscovery/).