AI-guided crispr-based TB drug-resistance test with optimized crrnas and engineered CAS enzymes

The National Institute of Allergy and Infectious Diseases awarded The Trustees of Princeton University $159,560 on November 28, 2026, under the Allergy and Infectious Diseases Research program (CFDA 93.855) to develop an AI-guided CRISPR-based diagnostic platform for detecting drug-resistant tuberculosis.

The research advances a CAS13 enzyme-based diagnostic system to detect drug-resistant TB with high accuracy. In the first objective, the recipient will apply a transformer-based machine learning model to decode sequence rules governing CAS13's activity and specificity. The work includes generating approximately 50,000 CRRNA/target pairs via CRISPR-based assays using the mCarmen microfluidic platform, which can test roughly 10,000 reactions simultaneously, and training the model to recognize sequence patterns including mismatches and insertions. The recipient will benchmark against state-of-the-art models including ADAPT and BADGERS. In the second objective, the recipient will use structure-guided protein engineering to create high-fidelity CAS13 variants through computational analysis of enzyme 3D structures via AlphaFold, construct a library of CAS13 mutants, and screen thousands of variants in parallel to identify improved enzymes with higher specificity (including single-nucleotide resolution) and maintained or improved sensitivity compared to wild-type CAS13.

Performance occurs at Princeton, New Jersey. The award period runs through June 30, 2029.

Abstract

Tuberculosis (TB), especially in drug-resistant forms, urgently requires faster and more reliable diagnostic tools. This project will advance a crispr-based diagnostic platform using CAS13 (an RNA-targeting enzyme) to detect drug-resistant TB with high accuracy. The broad objective is to decode the sequence rules that govern CAS13's activity and to engineer enhanced CAS13 enzymes for more precise and sensitive RNA detection. For this overarching goal, i will leverage a cutting-edge high-throughput microfluidic platform, mcarmen (microfluidic combinatorial arrayed reactions for multiplexed evaluation of nucleic acids), that is capable of testing ~10,000 reactions at once. In aim 1, i will apply artificial intelligence-specifically a transformer-based machine learning model-to learn how the sequences of crispr guide rnas and their RNA targets affect CAS13's activity and specificity. I will generate a large dataset of rifampicin-resistant target and crrnas (~50,000 crrna/target pairs) via crispr-based assays based on mcarmen, and train the transformer model to recognize the sequence patterns (including mismatches or insertions). I will benchmark against state-of-art models including adapt (activity-informed design with all-inclusive patrolling of targets) and badgers (building artificial diagnostic guides by exploring regions of sequences). In aim 2, i will use structure-guided protein engineering to create high-fidelity, high-activity CAS13 variants. By analyzing 3D structures (using computational models such as alphafold) of CAS13, i will identify key regions of the enzyme to modify. I will then construct a library of CAS13 mutants and screen thousands of variants in parallel to find improved enzymes that have higher specificity (e.g., single-nucleotide resolution) without sacrificing sensitivity or with better sensitivity than wild-type CAS13. Integrating them together, this work will result in a predictive tool for designing effective crispr diagnostics and new CAS13 enzyme variants that greatly improve test accuracy. While the main goal is to develop a point-of-care testing tool for detecting drug-resistant TB, more broadly, i envision that the insights and technologies developed will extend to other infectious diseases and diverse crispr-based applications including gene editing, RNA imaging and beyond. This fellowship will provide a critical training opportunity for becoming an independent researcher, focusing on the development of next-generation diagnostic tools. The expertise of my sponsor, coupled with the abundant resources available at princeton university (e.g., core facilities, high-performance computing centers, world- class researchers, and a mentoring committee), will allow me to substantially expand my postdoc training-from learning new skills (e.g., protein engineering, fundamental understanding of crispr systems, and deep learning) to career development through departmental and university-wide programs.