UT Arlington’s Huang Secures $3.1M NIH Grant to Advance AI-Driven Antibody Design via Protein Contact Map Prediction

Junzhou Huang, Jenkins Garrett Endowed Professor in the Department of Computer Science and Engineering at The University of Texas at Arlington, has secured a $3.1 million R01 grant from the National Institutes of Health to develop artificial intelligence methodologies for accelerated antibody drug discovery. The research focuses on computationally predicting antibody binding interactions to streamline the early stages of drug development, a process traditionally protracted and costly. Building upon established work in protein structure prediction, Huang’s team recently achieved sixth place overall in an international AI challenge, competing against institutions including Google DeepMind and the University of Washington, and notably ranked first in the protein contact map prediction track – a metric assessing the accurate prediction of inter-residue protein interactions. This success has fostered a collaborative research partnership with Tao Wang at UT Southwestern, aimed at significantly reducing the time and financial investment currently required to bring novel treatments for infectious diseases and autoimmune disorders to market.

Accelerating Antibody Drug Discovery

Junzhou Huang, Jenkins Garrett Endowed Professor in the Department of Computer Science and Engineering at The University of Texas at Arlington, has secured a $3.1 million R01 grant from the National Institutes of Health to propel advancements in artificial intelligence applied to antibody drug discovery. This research endeavours to substantially reduce the protracted and costly drug development timeline – a process conventionally exceeding a decade and demanding billions of dollars in investment to deliver a viable treatment. The core focus lies in automating and refining the initial phases of drug development, specifically antibody design, through computational prediction of binding interactions. Antibodies, immunoglobulin proteins produced by the immune system, function by recognising and neutralising foreign entities, including viruses and bacteria; therefore, their precise design constitutes a critical step in developing therapies for infectious diseases and autoimmune conditions.

Huang’s approach centres on in silico modelling – utilising computational methods to simulate biological processes – to predict the affinity and specificity of antibody-antigen interactions. This circumvents the need for extensive and time-consuming laboratory-based screening of antibody candidates. The methodology leverages advances in machine learning, specifically deep neural networks, trained on vast datasets of protein structures and binding affinities. A key challenge in this is the prediction of the ‘contact map’ – a binary map indicating which amino acid residues within the antibody and its target antigen are in close proximity, crucial for determining the strength and specificity of the interaction. The team’.s recent success in an international AI challenge, where they achieved sixth place overall in protein structure prediction and first in the protein contact map prediction track, demonstrates their proficiency in this computationally intensive field, competing against institutions including Google DeepMind and the University of Washington. This achievement validates the efficacy of their algorithms and provides a robust foundation for the current research.
This project builds upon Huang’s established expertise in protein structure prediction, a field heavily reliant on algorithms like Rosetta and AlphaFold, and is now extending this capability to the specific context of antibody drug discovery. The research is further strengthened through a new collaboration with Tao Wang at UT Southwestern, leveraging his expertise in structural biology and immunology. The ultimate aim is to establish a predictive framework that can efficiently identify antibody sequences with desirable properties, significantly streamlining the early stages of drug development and reducing associated risks and costs. The work has the potential to accelerate the medical response to future pandemics by enabling the rapid identification of antibodies that can neutralise emerging pathogens.

Leveraging AI for Protein Prediction

Junzhou Huang, Jenkins Garrett Endowed Professor in the Department of Computer Science and Engineering at The University of Texas at Arlington, has secured a $3.1 million R01 grant from the National Institutes of Health to pioneer the application of artificial intelligence to antibody drug discovery. This research endeavours to substantially curtail the protracted and costly drug development process, conventionally exceeding a decade and requiring billions of dollars in investment. The core strategy focuses on automating and refining the initial phases of drug development, specifically antibody design, through computational prediction of binding interactions – a methodology poised to accelerate the pipeline and mitigate associated financial and logistical burdens. Antibodies, as proteins produced by the immune system to neutralise foreign substances including viruses, represent a foundational element in the development of treatments for both infectious diseases and autoimmune disorders.

The project’s methodological foundation rests upon advances in machine learning, particularly deep neural networks trained on extensive datasets encompassing protein structures and binding affinities. A critical aspect of this endeavour lies in the accurate prediction of the ‘contact map’ – a binary representation denoting the proximity of amino acid residues within both the antibody and its target antigen. This contact map is fundamental to determining the strength and specificity of the antibody-antigen interaction, and therefore, the efficacy of the potential therapeutic. The team’s recent performance in an international AI challenge, achieving sixth place overall in protein structure prediction and first in the protein contact map prediction track, validates the robustness of their algorithms against those developed by leading institutions such as Google DeepMind and the University of Washington. This success demonstrates their proficiency in computationally intensive techniques like homology modelling, ab initio protein structure prediction, and the use of threading algorithms.

Huang’s established expertise in protein structure prediction, building upon algorithms such as Rosetta and AlphaFold, is now being extended specifically to the context of antibody drug discovery. The research is further bolstered by a new collaborative partnership with Tao Wang at UT Southwestern, who brings expertise in structural biology and immunology. This collaboration will facilitate the validation of computationally predicted antibody structures and binding affinities through experimental techniques such as X-ray crystallography and surface plasmon resonance. The ultimate objective is to establish a predictive framework capable of efficiently identifying antibody sequences possessing desirable characteristics, thereby streamlining the early stages of drug development and reducing associated risks and costs. This work holds significant potential to accelerate the medical response to future pandemics by enabling the rapid identification of antibodies capable of neutralising emerging pathogens, a capability of paramount importance in global health security.