AI Feedback Loop for HTP Enzyme Screening

CD Biosynsis is a leader in using artificial intelligence to advance enzyme engineering. With a professional team of scientists, we have long been dedicated to providing researchers with targeted enzyme evolution and design services. As artificial intelligence has been widely used in various fields in recent years, our EnzymoGenius^TM platform also applies artificial intelligence in the high-throughput enzyme screening process to accelerate the discovery of enzyme mutants.

Overview

In the process of AI model training, on the one hand, the AI model needs to output the corresponding results in accordance with the prescribed algorithm; on the other hand, the developer needs to judge the results output from the AI model and provide feedback to the AI model with the results of the judgment. This kind of feedback loop is an important step in the AI training process. It is based on feedback from developers that enables AI to make deep learning and model improvements in the direction we want it to go. Here are two common kinds of feedback loops. Firstly, the positive feedback loop, which refers to that developers give positive feedback to the AI model to affirm the correctness of its outputs. Secondly, the negative feedback loop refers to developers giving negative feedback on the incorrect output of AI model. Generally, positive and negative feedback loops are often used together to provide feedback to the AI model and guide the AI model in deep learning.

Fig. 1 Closed-loop workflow for computational autonomous molecular design. (Joshi RP, et al., 2021)

Important Parameters for AI Feedback Loop

• Accuracy
  Accuracy is the percentage of correctly predicted outcomes in the full sample.
• Precision
  Precision is the probability that the sample is actually positive out of all the samples predicted to be positive.
• Recall
  Recall is the probability of being predicted positive in the original sample.

With a professional scientific team, CD Biosynsis has been dedicated to providing researchers with enzyme directed evolution and design services. Combined with the advantages of Artificial Intelligence, we use AI to guide enzyme variant discovery and screen for ideal enzyme variants. If you are interested in the exclusive customization service for enzymes, please do not hesitate to contact us.

Reference

1. Joshi, RP.; Kumar, N. Artificial Intelligence for Autonomous Molecular Design: A Perspective. Molecules. 2021, 26(22):6761.