Home - skurvits/NN-for-virus-prediction GitHub Wiki

Welcome to the NN-for-virus-prediction wiki!

Link to the original article: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0222271

This is a practical example, but there is room for a theoretical study: Generate artificial sequences and labels for which we know exactly the generation rules -- e.g. one of the classes has the subword “ATTC” more often than the other class, some patterns only ever appear in one class (e.g. specific promoter regions) and so on. Explore what network architectures are able to learn on each type of data and which ones are not. The goal is to create a handbook for NNets for DNA analysis - if a geneticist has a dataset and has presumptions of what are the type of differences that separate the classes (ill/healthy; bacteria/virus), he/she can pick a network type based on your work on what works for which kind of data. (can be extended to study RNNs applicability too)