Tasks and Enhancements

[enryH]

bash scripts

• bash scripts for search bin/run_search.sh uses up to 12 CPUs. Could be extended by re-organizing the created in ./blobs by ./bin/create_blobs.sh into one folder continuously enumerated blob_001 ... blob_720. Then ./bin/run_search.sh could assign any to be create process a fraction of the blobs as inputs: ca. 720 blobs divided by # process
• speed up creation of blobs by multiprocessing the .smiles input files, or splitting it up into several files, starting more parallel process in ./bin/run_search.sh

Keyword Argument

• add keyword argument for fingerprints
• add keyword argument for similarity metric

Add Tests

• add public available .smiles files of a few thousand lines for testing from somewhere
• Write Unit-Tests for Functions
• Write Procedural Test for Scripts

Python Multiprocessing

• repair multiprocessing in python
• reading from file: Performance by processing line by line (no) vs chunk by chunck in python

Logging

• add logging of runs ?

Use zipped files

• create_blobs.py currently reads smiles-files, but the original data is zipped. Check if reading directly from zipped files leaves performance similar, see modular zipfile with zipfile.open method.
• blobs created by create_blobs.py are very large (500-600 GB) for the full Enamine REAL dataset. Check if compression leaves performance similar, e.g. using zipfiles.

To check

• check fragmentation algorithm on results?

Benchmarking to other solutions

Solution using chemfp

• compare to chemfp

Solution using rdkit functionality

• 1. Create Postgres Database Cartridge
• 2. Use SimilarityScreener