Computer Science Department
Using Neural Networks to Predict Pathogenicity of Amino Acid Mutations
My project is more research in nature and has a goal of using artificial intelligence techniques to determine if a mutation of an amino acid sequence within a Lysosomal Storage Disease (LSD) enzyme causes it to be pathogenic. This project implements back propagation, a supervised learning algorithm, written in Java. It will read in a data file containing information about each mutation (the attributes) such as its hydrophobicity score and accessible surface area, which will be obtained from the protein databank and other biochemistry databases. After the program reads in the input file, it will normalize the data and train the network. The output will be binary of whether the mutation is pathogenic or not (the classification). The network will then be tested using ten-fold cross-validation. The program will be tested using various learning rates in order to compare and minimize error rates.