# CustomKinFragLib: Filtering the Kinase-Focused Fragmentation Library

## TL;DR

CustomKinFragLib is a tool that filters a large library of kinase-targeting fragments to create a smaller, more practical set for drug design.

## Contribution

Introduces a customizable pipeline to reduce and filter a kinase fragment library for drug-like and synthesizable fragments.

## Key findings

- The pipeline reduces the original 9131 fragments to 837 while maintaining diversity and drug-like properties.
- Filters include synthetic accessibility, commercial availability, and removal of unwanted substructures.
- The resulting fragment set is suitable for downstream drug design workflows.

## Abstract

Protein kinases play a crucial role in key regulatory
cell processes
and are known to be dysregulated in diseases such as cancer and autoimmune
disorders. Hence, protein kinases represent a vital drug target class.
To meet the challenge of designing novel kinase inhibitors, fragment-based
drug discovery (FBDD) has already shown great promise. The kinase-specific
fragment library KinFragLib is a data-driven FBDD approach providing
a powerful subpocket-specific framework for creating potentially feasible
kinase inhibitors through subpocket-guided enumeration and combination
of fragments. However, traversing the whole recombination space is
computationally infeasible. Here, we introduce CustomKinFragLib, a
curation-focused and user-oriented pipeline that builds on the existing
KinFragLib framework. Building on the underlying fragmentation methodology,
CustomKinFragLib contributes a systematic post hoc reduction and filtering
strategy to generate a smaller, more tractable, and synthesis-friendly
fragment set. The pipeline integrates literature-derived drug-relevant
filters, including assessments of synthetic accessibility, matching
to commercially available building blocks, and availability of retrosynthetic
pathways. It also considers molecular properties often associated
with drug-likeness and removes fragments containing unwanted substructures.
Applying these curated filters reduces the original KinFragLib from
9131 to 837 fragments while retaining diverse fragments with drug-like
properties and high synthetic tractability, and providing a practical
fragment set suitable for downstream design workflows. Our pipeline
is easily customizable, allowing for modifications or exclusion of
filters based on the user’s preferences. The code and data
set are available at https://github.com/volkamerlab/KinFragLib.