BACKGROUND: Learning the representation of molecules is crucial for drug discovery but is often hindered by the scarcity of labeled experimental data, which limits the performance of supervised machine learning models. While self-supervised learning (SSL) offers a solution by leveraging vast unlabeled chemical databases, many existing methods focus on learning from either local structural information or global molecular properties, but not both simultaneously. We introduce MoCL-GAT, a novel contrastive and transfer learning-based SSL framework that addresses this gap by simultaneously learning from two complementary objectives. It combines a local contrastive task on molecular subgraphs to capture fine-grained chemical environments with a global predictive task to learn holistic molecular descriptors. This dual-objective approach, powered by a Graph Attention Network, is designed to create more robust, versatile, and transferable molecular representations. RESULTS: Pre-trained on 1.9 million compounds, MoCL-GAT was fine-tuned on diverse benchmarks. It achieved state-of-the-art performance on molecular property prediction tasks, with an AUROC of 0.928 on BBBP and 0.768 on SIDER, and top-ranking RMSEs of 0.570 for ESOL and 1.818 for FreeSolv. Critically, fine-tuned models consistently and significantly outperformed models trained from scratch, confirming the value of pre-training. CONCLUSIONS: These results validate that MoCL-GAT's dual-objective approach learns highly effective and transferable representations, enabling more accurate and data-efficient predictions for key cheminformatics challenges. The source code for MoCL-GAT is publicly available on Zenodo at https://doi.org/10.5281/zenodo.16927285 .

Foundation models have shown remarkable success across scientific domains, yet their impact in chemistry remains limited due to the absence of diverse, large-scale, high-quality datasets that reflect the field's multifaceted nature. We present the ChemPile, an open dataset containing over 75 billion tokens of curated chemical data, specifically built for training and evaluating general-purpose models in the chemical sciences. The dataset mirrors the human learning journey through chemistry---from educational foundations to specialized expertise---spanning multiple modalities and content types including structured data in diverse chemical representations (SMILES, SELFIES, IUPAC names, InChI, molecular renderings), scientific and educational text, executable code, and chemical images. ChemPile integrates foundational knowledge (textbooks, lecture notes), specialized expertise (scientific articles and language-interfaced data), visual understanding (molecular structures, diagrams), and advanced reasoning (problem-solving traces and code)---mirroring how human chemists develop expertise through diverse learning materials and experiences. Constructed through hundreds of hours of expert curation, the ChemPile captures both foundational concepts and domain-specific complexity. We provide standardized training, validation, and test splits, enabling robust benchmarking. ChemPile is openly released via HuggingFace with a consistent API, permissive license, and detailed documentation. We hope the ChemPile will serve as a catalyst for chemical AI, enabling the development of the next generation of chemical foundation models.

As the number of protein sequences increases in biological databases, computational methods are required to provide accurate functional annotation with high coverage. Although several machine learning methods have been proposed for this purpose, there are still two main issues: (i) construction of reliable positive and negative training and validation datasets, and (ii) fair evaluation of their performances based on predefined experimental settings. To address these issues, we have developed ProFAB: Open Protein Functional Annotation Benchmark, which is a platform providing an infrastructure for a fair comparison of protein function prediction methods. ProFAB provides filtered and preprocessed protein annotation datasets and enables the training and evaluation of function prediction methods via several options. We believe that ProFAB will be useful for both computational and experimental researchers by enabling the utilization of ready-to-use datasets and machine learning algorithms for protein function prediction based on Gene Ontology terms and Enzyme Commission numbers. ProFAB is available at https://github.com/kansil/ProFAB and https://profab.kansil.org.

MOTIVATION: To provide high quality, computationally tractable annotation of binding sites for biologically relevant (cognate) ligands in UniProtKB using the chemical ontology ChEBI (Chemical Entities of Biological Interest), to better support efforts to study and predict functionally relevant interactions between protein sequences and structures and small molecule ligands. RESULTS: We structured the data model for cognate ligand binding site annotations in UniProtKB and performed a complete reannotation of all cognate ligand binding sites using stable unique identifiers from ChEBI, which we now use as the reference vocabulary for all such annotations. We developed improved search and query facilities for cognate ligands in the UniProt website, REST API and SPARQL endpoint that leverage the chemical structure data, nomenclature and classification that ChEBI provides. AVAILABILITY AND IMPLEMENTATION: Binding site annotations for cognate ligands described using ChEBI are available for UniProtKB protein sequence records in several formats (text, XML and RDF) and are freely available to query and download through the UniProt website (www.uniprot.org), REST API (www.uniprot.org/help/api), SPARQL endpoint (sparql.uniprot.org/) and FTP site (https://ftp.uniprot.org/pub/databases/uniprot/). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Utilizing AI-driven approaches for drug-target interaction (DTI) prediction require large volumes of training data which are not available for the majority of target proteins. In this study, we investigate the use of deep transfer learning for the prediction of interactions between drug candidate compounds and understudied target proteins with scarce training data. The idea here is to first train a deep neural network classifier with a generalized source training dataset of large size and then to reuse this pre-trained neural network as an initial configuration for re-training/fine-tuning purposes with a small-sized specialized target training dataset. To explore this idea, we selected six protein families that have critical importance in biomedicine: kinases, G-protein-coupled receptors (GPCRs), ion channels, nuclear receptors, proteases, and transporters. In two independent experiments, the protein families of transporters and nuclear receptors were individually set as the target datasets, while the remaining five families were used as the source datasets. Several size-based target family training datasets were formed in a controlled manner to assess the benefit provided by the transfer learning approach. RESULTS: Here, we present a systematic evaluation of our approach by pre-training a feed-forward neural network with source training datasets and applying different modes of transfer learning from the pre-trained source network to a target dataset. The performance of deep transfer learning is evaluated and compared with that of training the same deep neural network from scratch. We found that when the training dataset contains fewer than 100 compounds, transfer learning outperforms the conventional strategy of training the system from scratch, suggesting that transfer learning is advantageous for predicting binders to under-studied targets. AVAILABILITY AND IMPLEMENTATION: The source code and datasets are available at https://github.com/cansyl/TransferLearning4DTI. Our web-based service containing the ready-to-use pre-trained models is accessible at https://tl4dti.kansil.org.

SUMMARY: Accurate prediction of the subcellular locations (SLs) of proteins is a critical topic in protein science. In this study, we present SLPred, an ensemble-based multi-view and multi-label protein subcellular localization prediction tool. For a query protein sequence, SLPred provides predictions for nine main SLs using independent machine-learning models trained for each location. We used UniProtKB/Swiss-Prot human protein entries and their curated SL annotations as our source data. We connected all disjoint terms in the UniProt SL hierarchy based on the corresponding term relationships in the cellular component category of Gene Ontology and constructed a training dataset that is both reliable and large scale using the re-organized hierarchy. We tested SLPred on multiple benchmarking datasets including our-in house sets and compared its performance against six state-of-the-art methods. Results indicated that SLPred outperforms other tools in the majority of cases. AVAILABILITY AND IMPLEMENTATION: SLPred is available both as an open-access and user-friendly web-server (https://slpred.kansil.org) and a stand-alone tool (https://github.com/kansil/SLPred). All datasets used in this study are also available at https://slpred.kansil.org. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Identification of interactions between bioactive small molecules and target proteins is crucial for novel drug discovery, drug repurposing and uncovering off-target effects. Due to the tremendous size of the chemical space, experimental bioactivity screening efforts require the aid of computational approaches. Although deep learning models have been successful in predicting bioactive compounds, effective and comprehensive featurization of proteins, to be given as input to deep neural networks, remains a challenge. RESULTS: Here, we present a novel protein featurization approach to be used in deep learning-based compound-target protein binding affinity prediction. In the proposed method, multiple types of protein features such as sequence, structural, evolutionary and physicochemical properties are incorporated within multiple 2D vectors, which is then fed to state-of-the-art pairwise input hybrid deep neural networks to predict the real-valued compound-target protein interactions. The method adopts the proteochemometric approach, where both the compound and target protein features are used at the input level to model their interaction. The whole system is called MDeePred and it is a new method to be used for the purposes of computational drug discovery and repositioning. We evaluated MDeePred on well-known benchmark datasets and compared its performance with the state-of-the-art methods. We also performed in vitro comparative analysis of MDeePred predictions with selected kinase inhibitors' action on cancer cells. MDeePred is a scalable method with sufficiently high predictive performance. The featurization approach proposed here can also be utilized for other protein-related predictive tasks. AVAILABILITY AND IMPLEMENTATION: The source code, datasets, additional information and user instructions of MDeePred are available at https://github.com/cansyl/MDeePred. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Predictive approaches such as virtual screening have been used in drug discovery with the objective of reducing developmental time and costs. Current machine learning and network-based approaches have issues related to generalization, usability, or model interpretability, especially due to the complexity of target proteins' structure/function, and bias in system training datasets. Here, we propose a new method "DRUIDom" (DRUg Interacting Domain prediction) to identify bio-interactions between drug candidate compounds and targets by utilizing the domain modularity of proteins, to overcome problems associated with current approaches. DRUIDom is composed of two methodological steps. First, ligands/compounds are statistically mapped to structural domains of their target proteins, with the aim of identifying their interactions. As such, other proteins containing the same mapped domain or domain pair become new candidate targets for the corresponding compounds. Next, a million-scale dataset of small molecule compounds, including those mapped to domains in the previous step, are clustered based on their molecular similarities, and their domain associations are propagated to other compounds within the same clusters. Experimentally verified bioactivity data points, obtained from public databases, are meticulously filtered to construct datasets of active/interacting and inactive/non-interacting drug/compound-target pairs (~2.9M data points), and used as training data for calculating parameters of compound-domain mappings, which led to 27,032 high-confidence associations between 250 domains and 8,165 compounds, and a finalized output of ~5 million new compound-protein interactions. DRUIDom is experimentally validated by syntheses and bioactivity analyses of compounds predicted to target LIM-kinase proteins, which play critical roles in the regulation of cell motility, cell cycle progression, and differentiation through actin filament dynamics. We showed that LIMK-inhibitor-2 and its derivatives significantly block the cancer cell migration through inhibition of LIMK phosphorylation and the downstream protein cofilin. One of the derivative compounds (LIMKi-2d) was identified as a promising candidate due to its action on resistant Mahlavu liver cancer cells. The results demonstrated that DRUIDom can be exploited to identify drug candidate compounds for intended targets and to predict new target proteins based on the defined compound-domain relationships. Datasets, results, and the source code of DRUIDom are fully-available at: https://github.com/cansyl/DRUIDom.

Bioinformatics (2010) doi: 10.1093/bioinformatics/btaa496 Co-author Tunca Doğan’s name was initially misspelled as Tunca Doğana. This has been corrected online.

SUMMARY: iBioProVis is an interactive tool for visual analysis of the compound bioactivity space in the context of target proteins, drugs and drug candidate compounds. iBioProVis tool takes target protein identifiers and, optionally, compound SMILES as input, and uses the state-of-the-art non-linear dimensionality reduction method t-Distributed Stochastic Neighbor Embedding (t-SNE) to plot the distribution of compounds embedded in a 2D map, based on the similarity of structural properties of compounds and in the context of compounds' cognate targets. Similar compounds, which are embedded to proximate points on the 2D map, may bind the same or similar target proteins. Thus, iBioProVis can be used to easily observe the structural distribution of one or two target proteins' known ligands on the 2D compound space, and to infer new binders to the same protein, or to infer new potential target(s) for a compound of interest, based on this distribution. Principal component analysis (PCA) projection of the input compounds is also provided, Hence the user can interactively observe the same compound or a group of selected compounds which is projected by both PCA and embedded by t-SNE. iBioProVis also provides detailed information about drugs and drug candidate compounds through cross-references to widely used and well-known databases, in the form of linked table views. Two use-case studies were demonstrated, one being on angiotensin-converting enzyme 2 (ACE2) protein which is Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Spike protein receptor. ACE2 binding compounds and seven antiviral drugs were closely embedded in which two of them have been under clinical trial for Coronavirus disease 19 (COVID-19). AVAILABILITY AND IMPLEMENTATION: iBioProVis and its carefully filtered dataset are available at https://ibpv.kansil.org/ for public use. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The identification of interactions between drugs/compounds and their targets is crucial for the development of new drugs. In vitro screening experiments (i.e. bioassays) are frequently used for this purpose; however, experimental approaches are insufficient to explore novel drug-target interactions, mainly because of feasibility problems, as they are labour intensive, costly and time consuming. A computational field known as 'virtual screening' (VS) has emerged in the past decades to aid experimental drug discovery studies by statistically estimating unknown bio-interactions between compounds and biological targets. These methods use the physico-chemical and structural properties of compounds and/or target proteins along with the experimentally verified bio-interaction information to generate predictive models. Lately, sophisticated machine learning techniques are applied in VS to elevate the predictive performance. The objective of this study is to examine and discuss the recent applications of machine learning techniques in VS, including deep learning, which became highly popular after giving rise to epochal developments in the fields of computer vision and natural language processing. The past 3 years have witnessed an unprecedented amount of research studies considering the application of deep learning in biomedicine, including computational drug discovery. In this review, we first describe the main instruments of VS methods, including compound and protein features (i.e. representations and descriptors), frequently used libraries and toolkits for VS, bioactivity databases and gold-standard data sets for system training and benchmarking. We subsequently review recent VS studies with a strong emphasis on deep learning applications. Finally, we discuss the present state of the field, including the current challenges and suggest future directions. We believe that this survey will provide insight to the researchers working in the field of computational drug discovery in terms of comprehending and developing novel bio-prediction methods.

BACKGROUND: The automated prediction of the enzymatic functions of uncharacterized proteins is a crucial topic in bioinformatics. Although several methods and tools have been proposed to classify enzymes, most of these studies are limited to specific functional classes and levels of the Enzyme Commission (EC) number hierarchy. Besides, most of the previous methods incorporated only a single input feature type, which limits the applicability to the wide functional space. Here, we proposed a novel enzymatic function prediction tool, ECPred, based on ensemble of machine learning classifiers. RESULTS: In ECPred, each EC number constituted an individual class and therefore, had an independent learning model. Enzyme vs. non-enzyme classification is incorporated into ECPred along with a hierarchical prediction approach exploiting the tree structure of the EC nomenclature. ECPred provides predictions for 858 EC numbers in total including 6 main classes, 55 subclass classes, 163 sub-subclass classes and 634 substrate classes. The proposed method is tested and compared with the state-of-the-art enzyme function prediction tools by using independent temporal hold-out and no-Pfam datasets constructed during this study. CONCLUSIONS: ECPred is presented both as a stand-alone and a web based tool to provide probabilistic enzymatic function predictions (at all five levels of EC) for uncharacterized protein sequences. Also, the datasets of this study will be a valuable resource for future benchmarking studies. ECPred is available for download, together with all of the datasets used in this study, at: https://github.com/cansyl/ECPred . ECPred webserver can be accessed through http://cansyl.metu.edu.tr/ECPred.html .

MOTIVATION: Similarity-based methods have been widely used in order to infer the properties of genes and gene products containing little or no experimental annotation. New approaches that overcome the limitations of methods that rely solely upon sequence similarity are attracting increased attention. One of these novel approaches is to use the organization of the structural domains in proteins. RESULTS: We propose a method for the automatic annotation of protein sequences in the UniProt Knowledgebase (UniProtKB) by comparing their domain architectures, classifying proteins based on the similarities and propagating functional annotation. The performance of this method was measured through a cross-validation analysis using the Gene Ontology (GO) annotation of a sub-set of UniProtKB/Swiss-Prot. The results demonstrate the effectiveness of this approach in detecting functional similarity with an average F-score: 0.85. We applied the method on nearly 55.3 million uncharacterized proteins in UniProtKB/TrEMBL resulted in 44 818 178 GO term predictions for 12 172 114 proteins. 22% of these predictions were for 2 812 016 previously non-annotated protein entries indicating the significance of the value added by this approach. AVAILABILITY AND IMPLEMENTATION: The results of the method are available at: ftp://ftp.ebi.ac.uk/pub/contrib/martin/DAAC/ CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.