Distance-as-function descriptor of binding pockets supports alignment-free pocket matching
- Rather than fixing one distance cutoff around the ligand, the method describes pockets as functions of distance, combining covariance-based structural descriptors with log-ratio-transformed elemental composition.
- Tested retrospectively on three public benchmarks (ligand discrimination, three-class pocket classification, pairwise pocket matching), using standard statistical learning and a Siamese network.
- Results were competitive with leading methods, with advantages depending on task; retrieval examples found pockets with related ligands despite low protein similarity.
💡 Points Based on the abstract
If you need to flag off-target pockets for a selectivity screen, an alignment-free matcher is the right tool, and the distance-as-function idea sidesteps the arbitrary cutoff problem that plagues fpocket-style descriptors. But three retrospective benchmarks say little about apo or induced-fit pockets, and I'd want a head-to-head against SiteEngine and graph-based comparison on my own target family before trusting it.
📄 Paper · Journal of Chemical Information and Modeling ↗ FUSED: A Functional Representation for Joint Structural and Elemental Analysis of Protein–Ligand Binding Sites