## Model Overview

We set up a model for facilitating the transformation of new chassis bacteria

We preliminarily assume that there is a simple linear relationship between the 64 components of this vector and the final result, and preliminarily confirm our hypothesis by comparing the difference between calculated value and measured value by solving sparse equations. Subsequently, we further strengthened our model by using the neural network model with stronger stability and more relevant parameters, and obtained an algorithm to predict the promoter strength.

We tested 25 promoters’ strength with firefly luciferase and another variable goes to promoter sequence gotten from pf5 genome in ncbi. As we know, house-keeping genes sometiomes express strongly and consistently. So we picked up 70 base pairs up from several genes of them as strong promoters in ^{[1]}

By combining three bases to encode DNA, solving a sparse system of equations,and strengthening the linear relationship between system input and promoter strength by neural network, a good correlation was found between the promoter sequence and promoter strength. It is through this mutually-informing modelling that we were able to figure out the relationship between promoter sequence and strength in P. fluorescence. This model informs us a promoter strength based on its specific sequence and it has also been verified by using other parts. This approach is also of valuable reference for future teams’ project.