Scientific programming course. Assignment 2 about multivariate statistics

Copywright This project is part of Scientific programming assignment developed by authors below which is released under MIT license. © 2011 Dara Akdag All Rights Reserved

Dara Sevkan Akdag. Maastricht university Systems Biology M.Sc.

The jupyter notebook can be executed in your web browser after jupyter notebook has been succesfully installed. Installation guide can be found at: http://jupyter.org/install The kernel used to execute the notebook is based on an anaconda distribution of python 3. Anaconda is a package handler which is great for installing libraries. https://conda.io/docs/user-guide/install/index.html

The descriptors matrix was computed based on variables from PubChemAid 624202. The study is called "QHTS Assay To Identify Small Molecule Activators Of BRCA1 Expression". Reference: National Center for Biotechnology Information. PubChem BioAssay Database; AID=624202, https://pubchem.ncbi.nlm.nih.gov/bioassay/624202 (accessed Oct. 5, 2018).

The pipeline of the notebook is split in 3 main parts.

• Data preparation (removing NaNs and making sure activity scores are associated with the correct descriptor entries)
• Data exploration (Boxplot, heatmap and PCA.)
• Data analysis (PLS with cross validation, Ridge regression, Lasso regression and elastic net).

Next step could be to apply forward feature selection. I did some preliminary feature selection but without any luck, so I removed it from the analysis.

The expected output can be viewed the the jupyter notebook which is uploaded. You should be able to get a descriptor matrix which contain the activity scores. And all analyses should be available once the appropriate libraries are installed.

Pandas http://pandas.pydata.org/pandas-docs/stable/api.html Numpy http://www.numpy.org/ Scikit learn packages: http://scikit-learn.org/stable/# Matplotlib https://matplotlib.org/ Seaborn https://seaborn.pydata.org/