Interdisciplinary techniques in protein binding prediction and crystal engineering

Abstract

This dissertation explores the integration of interdisciplinary methods such as advanced robotic automation, machine learning, and hybrid materials synthesis to dual protein engineering challenges: predicting protein-peptide binding specificity and the preparation of crystalline protein materials. The first chapter introduces a computational pipeline, PAbFold, based on AlphaFold2, designed to predict linear antibody epitopes from a given antigen sequence. This method provides a rapid and cost-effective alternative to traditional experimental techniques for epitope mapping, significantly lowering the financial barrier for laboratories. By accurately identifying binding sites on target proteins, PAbFold enhances the understanding of antibody-antigen interactions, facilitating the development of diagnostic and therapeutic antibodies in a more accessible manner. The second chapter presents an innovative approach to protein crystallization scale-up utilizing the Opentrons 2 liquid handling robot. This automation not only reduces manual labor and variability in crystallization experiments but also makes high-throughput crystallization more accessible to a broader range of laboratories by decreasing costs. Traditional high throughput protein crystallization liquid handling robots are priced around $75,000; the Opentrons 2 costs around $15,000. By employing Python scripts for precise control of the Opentrons 2, the study demonstrates successful crystallization of model and non-model proteins, highlighting the potential of automated systems in structural biochemistry to democratize access to high-quality protein crystals. The third chapter delves into the creation of hybrid materials by combining metal-organic frameworks (MOFs) with porous protein crystals. The research demonstrates the feasibility of embedding MOF domains within protein crystals, potentially opening new avenues for applications in catalysis, gas storage, and chemical warfare agent detoxification. By developing a new class of hybrid materials, this work contributes to making advanced structural biochemical research. Together, these chapters illustrate a modern interdisciplinary approach that embraces machine learning and automation in service of the engineering of peptide-binding proteins and crystalline protein materials. The integration of automation, computational predictions, and hybrid materials offers a promising path toward more efficient and innovative solutions in biochemical research, while significantly lowering the cost barriers, thereby increasing accessibility for researchers worldwide.