DEVELOPMENT OF STRUCTURAL INFORMATICS METHOD FOR BINDING PEPTIDES

In the continuous pursuit of advanced therapeutics, the field of bioinformatics has innovated tools that allow unprecedented control over the proteome, profoundly shaping our understanding and manipulation of biological domains. Computational approaches to protein design grapple with the intricacies of protein behavior, encompassing everything from interaction dynamics to stability challenges. Methods in structural bioinformatics for peptide design typically hinge on the datasets of structures that have statistics applied to ascertain the effectiveness of protein design and modulation. When dealing with proteins that are poorly resolved, disordered, or niche, this task usually falls to experts in structural biology and often requires significant laboratory resources.
This thesis discusses an automated pipeline, devised to integrate remote sequence homology, structural modeling, and binding simulations of peptides to disordered proteins. Significant design and testing underpin this pipeline, aiming to generate binding peptides to any sequence, sidestepping the absolute requirement for an expert or a tedious process to produce leads. The utility of this pipeline is assessed across a diverse set of protein systems to refine its methodology. With the recent rise of machine learning-driven predictive or generative models, we explore their potential when integrated with our pipeline in attempt to address challenges in the computation of peptide binder design.