Note: New software is being written using the pseudocode/basic abstraction which will ultimately yield new molecular descriptors that can be used for Machine Learning.
Introduction
Novel research idea in the Dr. Jacob D. Durrant Lab of Computational Drug Discovery at the University of Pittsburgh (Lab Page) that wanted to find new ways of extracting potentially significant features/descriptors of molecules by reutilizing and modifying code from NNScore 2.0. NNScore 2.0 is a neural network-based ligand-protein interaction prediction software that takes into account the many available descriptors/features of a Protein Data Bank (PDB) file to ultimately rank many possible ligands against a protein. This open-source software can be downloaded at SourceForge. The academic paper that describes the software in depth can be found here.
Documents in Folder
📜 ML_Features_NNScore2.0_Basic_Abstraction.txt
Contains basic structure of the code in a language agnostic manner while including all the different variables required. This is mainly surface-level detail abstraction of the necessary code to extract potentially meaningful molecular features.
📜 NNSCore2.0_Descriptors_Pseudocode.txt
Well-organized and final pseudocode document that lays out the actual analysis and reasons behind the code and the way/order in which it is executed. It gives clear explanation/instruction on what needs to be done at each point in such a way that the pseudocode can be taken and modified heavily to yield the same result as the original software.
📁 Python Files/
The original text files have been converted to Python files without any real modifications. Obviously, the Python files are not able to run. The only reason this has also been provided is for those that like syntax highlighting when viewing the pseudocode/abstraction. They can be found in the Python Copy folder.
Contact Information
Yogindra Raghav (YogiOnBioinformatics)