Protein domain-based prediction of drug/compound–target interactions and experimental validation on LIM kinases

Abstract

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.

How to approach machine learning-based prediction of drug/compound–target interactions

“…Considering algorithmic approaches, sequence-based protein representations can be grouped as conventional/ classical descriptors (or descriptor sets) and learned embeddings. Conventional descriptors are mostly modeldriven, meaning that they are generated by applying predefined rules and/or statistical calculations on sequences considering various molecular properties that include physicochemical [10][11][12], geometrical [13,14] and topological [12] characteristics of amino acids, as well as sequence composition [11,15], semantic similarities [16], functional characteristics/properties [17][18][19][20], and evolutionary relationships [13,21] of proteins. Learned protein embeddings (a.k.a.…”

Machine learning in computational modelling of membrane protein sequences and structures: From methodologies to applications

“…Given that membrane proteins are greatly involved in DTIs, as indicated above, we assume that these tools will also have an acceptable ability to predict DTIs when the drug target is a membrane protein, even though the precise ratio of membrane proteins vs. globular proteins was not made clear by these aforementioned studies in their training datasets. This has been suggested in the DRUIDom work [263] . However, compared with sequence-based prediction of DTIs, the field is greatly devoid of prediction tools developed based on 3D protein structures [264] .…”

A machine learning-integrated stepwise method to discover novel anti-obesity phytochemicals that antagonize the glucocorticoid receptor

“…Compounds with IC 50 lower than 10 μM were labeled as positives and compounds with IC 50 higher than 20 μM were labeled as negatives. Compounds with IC 50 between 10 and 20 μM were considered ambiguous and were therefore discarded in a similar manner to a previous work by Rifaioglu et al 41 and Doğan et al 42 From the two databases, a total of 2281 instances (2242 positives and 39 negatives) were obtained. Since there was a higher proportion of negatives—compounds with weak or no GR binding affinities—in the real-world data, we designed the training dataset to have a ten-fold greater proportion of negatives than positives.…”