In this year's iGEM competition, our team made significant contributions by developing a novel model tool capable of qualitative and quantitative two-step screening of peptide sequences with potential antimicrobial activity. Additionally, we made great strides in laboratory safety protocols and practical education initiatives.

Database Integration and Data Set Contribution

We systematically screened and integrated peptide sequences with potential antimicrobial activity from the NCBI database. By processing and standardizing this data, we created a high-quality antimicrobial peptide dataset, which can serve future researchers and greatly improve the efficiency of peptide sequence screening.

Development of Machine Learning Tools Based on DeBERTa Model

We utilized the DeBERTa model to perform both qualitative and quantitative screenings. This model predicts antimicrobial peptide activity and their MIC values against various fungi, significantly improving the screening process by providing comprehensive sequence feature analysis.

Experimental Verification and Result Sharing

We validated the accuracy of the model predictions by performing antifungal activity tests in the lab, where we determined the MIC values of peptides and analyzed the consistency between experimental and predicted results.

Troubleshooting and Improvement

During model development, we found that the positive data was shorter while negative data was longer. As a result, the model tended to prioritize long data. We adjusted the pre-training to improve accuracy for positive data sets, which could serve as a reference for other teams working on similar tasks.

Future Suggestions and Directions

We found that for specific fungal species, sequence prediction accuracy was limited. We suggest future research optimizes model parameters and expands training datasets to improve prediction precision.

Our MIC testing and hemolysis experiments in mice greatly supported the practical significance of our model. Animal ethics has always been a major focus for our team. Before conducting any experiments, we organized thorough lab safety training, contributing to our entire team's awareness of animal ethics and laboratory safety standards.

To raise public awareness about fungal infections and machine learning, we developed the interactive game "End-time Survival Manual 2024IGEM-SWUER". This game places participants in the role of researchers, allowing them to explore concepts related to fungal infections and machine learning in an engaging way.

We also hosted offline workshops, receiving highly positive feedback due to the game's educational value and social impact. In the future, we aim to translate this interactive manual into multiple languages to reach a broader audience.

You can download the interactive manual here: End-time Survival Manual 2024IGEM-SWUER