There has recently been a rapid progress in computational methods for determining protein targets of small molecule drugs, which will be termed as compound protein interaction (CPI). In this review, we comprehensively review topics related to computational prediction of CPI. Data for CPI has been accumulated and curated significantly both in quantity and quality. Computational methods have become powerful ever to analyze such complex the data. Thus, recent successes in the improved quality of CPI prediction are due to use of both sophisticated computational techniques and higher quality information in the databases. The goal of this article is to provide reviews of topics related to CPI, such as data, format, representation, to computational models, so that researchers can take full advantages of these resources to develop novel prediction methods. Chemical compounds and protein data from various resources were discussed in terms of data formats and encoding schemes. For the CPI methods, we grouped prediction methods into five categories from traditional machine learning techniques to state-of-the-art deep learning techniques. In closing, we discussed emerging machine learning topics to help both experimental and computational scientists leverage the current knowledge and strategies to develop more powerful and accurate CPI prediction methods.

Kinase inhibitors are a major category of drugs. Experimental panel assay protocols are routinely used as a standard procedure to evaluate the efficiency and selectivity of a drug candidate to target kinase. However, current kinase panel assays are time-consuming and expensive. In addition, the panel assay protocols neither provide insights on binding sites nor allow experiments on mutated sequences or newly-characterized kinases. Existing virtual screening or docking simulation technologies require extensive computational resources, thus it is not practical to use them for the panel of kinases. With rapid advances in machine learning and deep learning technologies, a number of DTI tools have been developed over the years. However, these methods are yet to achieve prediction accuracies at the level of practical use. In addition, the performances of current DTI tools vary significantly depending on test sets. In this case, an ensemble model can be used to improve and stabilize DTI prediction accuracies. In this work, we propose an ensemble model, EnsDTI-kinase, that integrates eight existing machine learning and deep learning models into a unified model deployed as a web-server. Upon submission of a compound SMILES string, potential target kinases are automatically predicted and evaluated on the web-server. Importantly, EnsDTI-kinase is a computational platform where newly developed DTI tools can be easily incorporated without modifying core components so that its DTI prediction quality can improve over time.Besides, many useful functionalities are provided on our platform for users to further investigate predicted DTI: it allows confidence experiments by changing the amino acid (AA) at a specific position in a kinase sequence, named in-silico mutagenesis, to investigate the effect of AA changes in binding affinity; it predicts kinase sequential regions where the query compound likely binds to by slidingly masking the sequence of selected kinases so that confidence in the predicted binding sites can be evaluated. Our model was evaluated in three experimental settings using four independent datasets, and showed accuracy of 0.82 compared to the average accuracy of 0.69 from five deep learning methods on the ChEMBL dataset. It achieved average selectivity of 0.95 within kinase families such as TK, CAMK and STE. For 8 out of 17 recent drugs, our model successfully predicted their interactions with 404 proteins at average accuracy of 0.82.

Recent contrastive learning methods have shown to be effective in various tasks, learning generalizable representations invariant to data augmentation thereby leading to state of the art performances. Regarding the multifaceted nature of large unlabeled data used in self-supervised learning while majority of real-word downstream tasks use single format of data, a multimodal framework that can train single modality to learn diverse perspectives from other modalities is an important challenge. In this paper, we propose TriCL (Triangular Contrastive Learning), a universal framework for trimodal contrastive learning. TriCL takes advantage of Triangular Area Loss, a novel intermodal contrastive loss that learns the angular geometry of the embedding space through simultaneously contrasting the area of positive and negative triplets. Systematic observation on embedding space in terms of alignment and uniformity showed that Triangular Area Loss can address the line-collapsing problem by discriminating modalities by angle. Our experimental results also demonstrate the outperformance of TriCL on downstream task of molecular property prediction which implies that the advantages of the embedding space indeed benefits the performance on downstream tasks.Preprint. Under review.