Machine Learning-based PPI Focused Library

Protein-protein interactions (PPI) can control a large number of cellular processes such as growth, cell survival, cell adhesion, signal transduction, apoptosis, host-pathogen interactions and immune regulation. Deregulations of PPIs are associated with many different physio-pathologies such as cancer development, infectious diseases, neurological disorders, inflammation and oxidative stress disorders. Therefore, modulation of PPIs by small molecule inhibitors is increasingly recognized as a therapeutic intervention strategy in disease biology and drug discovery research. The selection of PPIs as targets for new drugs offers other advantages in terms of better selectivity or lower off target specificity and lower chance of resistance compared with targeting a single protein or enzyme. A diversity of structure-based approaches, such as docking and molecular dynamics simulation, have been widely used to screen large libraries of compounds against target enzymes or PPI chaperone proteins.

BOC Sciences has developed a machine learning-based PPI focused library to provide 6,800 potential PPI modulators.

Figure 1. Flowchart depicting hierarchical design of generalized and family specific predictors for identification of modulators of PPI. (Gupta, P.; et al. 2021)

Machine Learning-based PPI Focused Library Design

BOC Sciences has developed a decision tree strategy based on a cross-validation protocol to predict which small-molecule compounds can target PPIs, an approach that balances the richness, sensitivity, and specificity of the learning dataset.

• We compare unique physicochemical characteristics of non-PPI and PPI inhibitors, and identify several descriptors showing the relevance of PPI binders in the specified range
• A shape-based descriptor for defining the radial distribution function of a collection of atoms in a spherical volume with the radius of 7 Å
• An unsaturation index directly related to the number of multiple bonds, including double, triple and aromatic bonds
• An average shape profile index of order 2 derived from the distance distribution of the geometric matrix
• Descriptors calculated by summing the atomic weights viewed for different angular scattering functions

Machine Learning-based PPI Focused Library Characteristics

BOC Sciences’ machine learning-based PPI focused library includes a diversity of therapeutically relevant compounds that are carefully selected from our proprietary collection of HTS compounds to meet the parameters listed in the table below.

Table1. The summary of the BOC Sciences Machine Learning-based PPI Focused Library characteristics

Features of Machine Learning-based PPI Focused Library

• All PAIN and reactive compounds are excluded from selection by internal filter applications
• Structurally diverse subset, with the option to favor hit discovery
• Structural analogs available for SAR studies
• All compounds are continually updated
• Compound cherry-picking service is available

Figure 2. Family wise distribution of non-redundant PPI modulators in the compiled non-redundant data set. (Gupta, P.; et al. 2021)

What We Deliver

• Delivered within 2 weeks in any customer-preferred format
• Powders, dry films or DMSO solutions formatted in vials, 96 or 384-well plates
• All compounds have a minimum purity of 90% assessed by ¹H NMR and HPLC
• Analytical data is provided

BOC Sciences provides professional, rapid and high-quality services of Machine Learning-based PPI Focused Library design at competitive prices for global customers. Personalized and customized services of Machine Learning-based PPI Focused Library design can satisfy any innovative scientific study demands. Our clients have direct access to our staff and prompt feedback to their inquiries. If you are interested in our services, please contact us immediately!