(404c) Area 15A Plenary Award - Developing a Strain and Manufacturing Ai Research Tool (SMART) for Synthetic Biology Knowledge Synthesis and Output Prediction

Biomanufacturing has a potential US market of over 30 billion dollars annually (2023 Government Accountability Office report). A primary challenge that needs urging solutions, however, is the high cost to develop cellular factories that meet commercially relevant performance. The high cost is attributed to the many rounds of design-build-test-learn (DBTL) cycles as well as high-risk bioprocess development towards scaling-up. Biological systems are complex and there are many important levers (e.g. genetic regulations, enzyme functions, cellular metabolism, and extracellular conditions) that need to be tuned to engineer a desired phenotype such as production outputs. Therefore, a holistic knowledgebase from strain technology to manufacturing development is essential. This project aims to automate information gathering from synthetic biology (SynBio) literature, to perform biomanufacturing knowledge synthesis, and to make the production outputs of engineered microbes more predictable. Specifically, Large Language Models (LLMs) under human supervisions can facilitate the collection and organization of relevant information from vast SynBio literatures. Leveraging on Alibaba Qwen 通义千问, a knowledge synthesis workflow termed Network for Knowledge Organization can extract relevant datasets from texts, followed by feature engineering, organizations of structured datasets, and knowledge visualization of relevant biomanufacturing concepts. This workflow is more informative and specific to summarize and distill synthetic biology information than Open AI GPT-4 zero-shot Q&A. The resulting knowledge and structured database can not only assist synthetic biology research design and education, but also support machine learning, transfer learning and metabolic models to enable high quality predictions of various SynBio processes. These predictions can provide broad guidelines for effective DBTL strategies and optimal bioprocess conditions. Based on the above concepts, a Strain and Manufacturing Ai Research Tool (SMART) is being developed to mine literature data for generalizable rules of engineering biology, uncover patterns in complex data sets, rapidly generate SynBio prototype, and offer preliminary economic assessments.