Overview
This year, our team aims to develop a modularized adhesive platform, CoPlat. Therefore we designed lots of biobricks and a Machine Learning model (ML model) to construct and optimize our CoPlat. In this case, we find some strong composite parts and the high credibility of our ML model. We tend to show these two useful tools: Composite Part and Software, to future iGEM teams, and hope these can help them.
Composite Part
In our project, we designed lots of composite parts including six natural adhesive recombinant proteins, four potential adhesive recombinant proteins, and two functional recombinant proteins. After conducting three functional tests: Flushing test, Viscosity test, and modified ELISA, we confirmed the powerful seventh composite parts: CsgA-Bamcp20k-1 (BBa_K4854016), CsgA-cp19k (BBa_K4854019), CsgA-Mrcp20k (BBa_K4854020), CsgA-ecpA (BBa_K4854022), CsgA-Nid1 (BBa_K4854023), CsgA-epd2 (BBa_K4854024), and CsgA-zig4 (BBa_K4854025). If you want to see more detailed descriptions, please click on our Result link.
Through that evidence, we not only proved the success of producing CoPlat but we also confirmed the accuracy of our classifier is real. So we further develop software to make future iGEM teams more easily use our ML model.
Software
We have designed a user-friendly machine learning classifier where you can enter the EntryID of a protein in the user interface to get its adhesive score. This classifier uses ESM-2 as the encoding method and SVM as the algorithm.
The future iGEM team can not only employ our software to evaluate the adhesive of proteins, but also access our open-source code on GitHub, and use it alongside other training data to convert it into a classifier for evaluating other properties. If you want to see more detailed descriptions, please click on our Software link.