Add fraction best ligands metric

[matthewtwarren]

Description

Implementation of Chris Bayly's fraction of best ligands ranking/classification metric.

This involves first computing the overlap coefficient - the fraction of N most-active experimental ligands that are in the N most-active predictions. This is implemented via the _compute_overlap_coefficient helper function, which relies on a 2D histogram constructed via the _create_2d_histogram function.

The fraction of best ligands is then calculated by averaging the overlap coefficients from N = 1 up to a given fraction of the ligands. The default is 0.5, i.e. up to num_ligands/2, rounded down. This is implemented in via compute_fraction_best_ligands.

Todos

• Implement helper functions to generate histogram and compute overlap coefficient
• Implement main function to compute metric
• Write tests

Status

• Testing
• Ready to go

[pep8speaks]

Hello @matthewtwarren! Thanks for opening this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file cinnabar/classification_metric.py:

Line 151:1: W293 blank line contains whitespace
Line 202:83: W291 trailing whitespace
Line 231:1: W293 blank line contains whitespace
Line 234:1: W293 blank line contains whitespace
Line 237:33: W292 no newline at end of file

[RiesBen]

HI @matthewtwarren ,
thanks for pushing this, I will have a look at it soon! :)

[jthorton]

Thanks for the great work on this @matthewtwarren this is now completed in #164 in which we list you as an author!