AI Guided Precise HTP Enzyme Library Screening

Finding new enzyme variants with a desired substrate range requires screening a large number of potential variants. Although in a typical in silico enzyme engineering workflow, thousands of variants can be scanned and multiple candidates collected for further screening or experimental validation. However, this is not the best strategy for finding variants. CD Biosynsis is an expert in using artificial intelligence to advance enzyme engineering. We use machine learning (ML) to guide precise HTP library screening of enzyme engineering in order to find the most suitable variants and customize new enzymes or optimize certain properties for you.

Overview of AI guided precise HTP library screening

The combinatorial space of enzyme sequences has astronomical possibilities. Under the approach of restriction mutagenesis and combinatorial search, multi-target objectives (e.g., simultaneous improvement of enzyme selectivity, solubility, and activity) have narrow plausibility. Although the proliferation of high-throughput experimental methods, including next-generation sequencing and automated screening. However, current high-throughput screening of enzyme variants, experimentally or in silico, in either case, screening the entire combinatorial space may not be the most cost-effective strategy for finding suitable candidate variants. This requires us to rethink parallel approaches to enzyme engineering. A good alternative might be artificial intelligence (AI) and machine learning (ML), by training some algorithms, such as neural networks, which learn to combine mutations and predict outcomes at lightning speed (ultra-high-throughput screening). A typical example is researchers train a neural network to learn the intricate synergic relationships needed to assess mutants resulting of the combination of two or more individual mutations. The main advantage of the presented deep learning strategy is the massive speed-up, evaluating 108 variants in less than 24 h, compared to traditional computational HTS screening methods. The trained AI model can predict the binding energy of unseen enzyme variants in 1.36 ms with high accuracy, enabling super high-throughput screening.

Category: General