hippo-db 0.3.0

Hit Interaction Profiling and Procurement Optimisation

HIPPO

🦛 Hit Interaction Profiling for Procurement Optimisation

A pipeline for optimally selecting building blocks to maximise interaction diversity of reaction products for follow-up screens at XChem.

Inputs

• Crystallographic data from a fragment screen (i.e. from Fragalysis)
• A set of follow-up compounds with associated catalogue building blocks
• A budget for building block procurement

Outputs

• Interaction fingerprints for each hit and compound
• Scores for interaction coverage of a given compound set
• Sankey diagram of fragment screen interaction conservation
• UMAP reduction of the interaction fingerprints into 1D and 2D interaction space
• Suggested building block sets that optimise interaction coverage and stay within the budget

Installation

Python dependencies:

• UMAP
• RDKit
• Pandas
• Numpy
• Plotly
• ASE

Install from python source:

• MPyTools
• MolParse
• HIPPO

Usage

import hippo as hp

# create the hippo
pipeline = hp.HIPPO('project_name')

# protein APO structure
pipeline.add_protein_reference(path=protein_reference)

# get the fragment screen data
pipeline.add_hit_metadata(path=metadata_csv)
pipeline.add_hit_directory(path=aligned_path)

# add elaborations/merges
pipeline.add_product_compounds('compounds', metadata_csv, compound_directory, compound_mol_pattern)

# make the interaction fingerprints
pipeline.generate_fingerprints()

# interaction coverage
pipeline.score_interaction_coverage()

# UMAP visualisation (not implemented yet)
pipeline.plot_umap()

# optimise building block selection
pipeline.suggest_building_blocks(budget=10_000)

# store the whole pipeline as a binary
pipeline.write_pickle(pickle_path)

# load an existing pickled pipeline:
pipeline = hp.HIPPO.from_pickle(pickle_path)