Prediction of siRNA activity for RNA interference

Description

The set of scripts included in this repository use the dataset of siRNA created by Heusken et. al (Heusken, Nature Biotechnology, 2005). The goal is to develop, using this dataset, a regression model that will predict the efficacy of a given siRNA molecule for RNA interference and a classification model that will predict whether a given siRNA molecule will have the desired potency for RNA interference or not.

Dataset

The dataset comprises of a set of guide siRNA sequences from mouse and human and their measured activity against targetted mRNA sequences

Approach

List of features

The features were calculated using the calc_features_v2 script.

Results (classification model)

Performance with all features and performance convergence

Top features and performance with a subset of 92 features

Results (regression model)

Performance, convergence and important features for regression model