Molecules based on their fingerprint similarity and protein–molecules interactions play a fundamental role in a variety of biological processes, here we have introduced ML-PDDT application suite, a novel web server based on machine learning algorithms. In silico assessment of protein receptor interactions with small ligands is now part of the standard pipeline for drug discovery, and numerous tools and protocols have been developed for this purpose. With the ML-PDDT web server, we propose a new approach to facilitate access to small molecule identification for non-specialists. The ML-PDDT online application suite integrates 3 modules: 1. molecules screening based on their fingerprint similarity, 2. ligand based similarities search and 3. Target interactions, drug-target search, and drug-target periodization based on a machine learning approach. This application does not require advanced computer knowledge, and it works without the installation of any programs with the exception of a common web browser. The ML-PDDT process is fast and consumes about an average 2–5 min for a run. The ML-PDDT web server is available at https://cdri.res.in/mlpddt/ or https://mlpddt.streamlit.app. Moreover, the limitations of access and writing permissions on particular computer equipment make the installation of these programs even more complex and laborious. This is particularly the case in a pedagogical context (e.g., practical work), where the use of different Unix command line programs leads to a waste of time for students and supervisors. In our experience, most of the student issues come from a bad understanding of the command line environment. In the worst case, this distracts the students from the educational purpose of the session. With the ML-PDDT web server, we propose a new approach to facilitate access to small-molecule similarities search for non-specialists, including students. ML-PDDT can be seen as interfacing a docking library (CHEMBL, ZINC and Natural Drugs) interactively through a web browser. Each ML-PDDT run is in fact an interactive processing session. Communication between the browser and server is in real-time and bidirectional.
Module 1 (Omega-X)
Using ML-PDDT, we get familiar with different approaches to encode (descriptors, fingerprints) and compare (similarity measures) molecules. Furthermore, we perform a ligand based virtual screening in the form of a similarity search for the specific targets against OMEGA-X datasets of target-tested molecules from the ChEMBL & ZINC database filtered by Lipinski’s rule of five. Due to larger available data sources, machine learning (ML) gained momentum in drug discovery and especially in ligand-based virtual screening. In this OMEGA-X, how to use different supervised ML algorithms to predict the activity of novel compounds against our targets.
Module 2 (Alpha-X)
The ML-PDDT module integrates molecules screening from OMEGA-X based on their similarity based on a machine learning approach.
Module 3 (Beta-X)
The ML-PDDT also introduced the separate module (Beta-X) that clarifies direct and indirect partnerships associated with biological interactions with possible drug interactions. This novel approach in collects information using an improved algorithm by transferring interactions between more than 110000+ entries, allowing statistical analysis with the automated background for the given inputs for functional enrichment.