TOURIST: Octant-Based Spherical Harmonic Descriptors for Predicting Inter-Chain Contacts from Unbound Protein Structures

Predicting which residue pairs form contacts across a protein-protein interface from unbound monomer structures remains a fundamental challenge in computational structural biology. We present TOURIST (Three-dimensional Octant-based Unified Interface Representation via Spherical Harmonics), a geometric descriptor that encodes the local spatial distribution of neighboring atoms around each surface residue as a compact spherical harmonic (SH) signature. The method constructs a local coordinate frame at each residue's C-alpha atom with the \(z\)-axis aligned to the surface normal, partitions the surrounding volume into eight octants, and expands the atom distribution as real spherical harmonics up to \(l_{\mathrm{max}} = 4\), yielding a 25-dimensional \(L_2\)-normalized descriptor. Cross-chain complementarity is scored via a dot product with \((-1)^l\) parity, encoding the anti-parallel surface normal assumption characteristic of lock-and-key interfaces. Evaluated on 145 complexes from the Protein-Protein Docking Benchmark 5 (BM5), TOURIST achieves a mean per-complex AUC of 0.733 for predicting inter-chain residue contacts from unbound structures, significantly outperforming all structural baselines including SASA product (0.649), hydrophobicity product (0.524), charge attraction (0.505), and a supervised combined logistic regression model (0.653; leave-one-out cross-validation). Wilcoxon paired tests confirm the difference is highly significant (\(p < 0.0001\)), with TOURIST winning in 102 of 145 complexes. The method performs substantially better on complexes exhibiting lock-and-key geometry (mean normal dot product \(< -0.30\): AUC \(= 0.765\), \(n = 114\)) than on non-lock-and-key complexes (AUC \(= 0.562\), \(n = 31\)). Bound and unbound performance are nearly identical (0.722 vs 0.733), indicating robustness to conformational change upon binding. We characterize the scope of TOURIST through a pose discrimination experiment: the descriptor does not distinguish near-native from non-native docking decoys (mean AUC \(= 0.474\) on ZDOCK decoy sets), confirming that it captures pairwise interface complementarity rather than global pose quality. This defines TOURIST as a contact prediction method for unbound structures rather than a docking scoring function. Nevertheless, the unbound contact prediction performance demonstrates that geometric shape complementarity, encoded as octant-based SH descriptors, captures structural information beyond surface exposure and physicochemical properties alone. A companion web application provides interactive visualization of lock-and-key complementarity and geometric docking of unbound protein structures.

Authors

Publication Details

Journal
Preprints.org
Published
2026-11-13
DOI
https://doi.org/10.20944/preprints202611.1372.v1
Primary Topic
Protein Structure and Dynamics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

TOURIST: Octant-Based Spherical Harmonic Descriptors for Predicting Inter-Chain Contacts from Unbound Protein Structures

Nicholas Harkiolakis
1 links
Preprints.org
Protein Structure and Dynamics
preprint

TOURIST: Octant-Based Spherical Harmonic Descriptors for Predicting Inter-Chain Contacts from Unbound Protein Structures

Nicholas Harkiolakis
preprint en

Abstract

Predicting which residue pairs form contacts across a protein-protein interface from unbound monomer structures remains a fundamental challenge in computational structural biology. We present TOURIST (Three-dimensional Octant-based Unified Interface Representation via Spherical Harmonics), a geometric descriptor that encodes the local spatial distribution of neighboring atoms around each surface residue as a compact spherical harmonic (SH) signature. The method constructs a local coordinate frame at each residue's C-alpha atom with the \(z\)-axis aligned to the surface normal, partitions the surrounding volume into eight octants, and expands the atom distribution as real spherical harmonics up to \(l_{\mathrm{max}} = 4\), yielding a 25-dimensional \(L_2\)-normalized descriptor. Cross-chain complementarity is scored via a dot product with \((-1)^l\) parity, encoding the anti-parallel surface normal assumption characteristic of lock-and-key interfaces. Evaluated on 145 complexes from the Protein-Protein Docking Benchmark 5 (BM5), TOURIST achieves a mean per-complex AUC of 0.733 for predicting inter-chain residue contacts from unbound structures, significantly outperforming all structural baselines including SASA product (0.649), hydrophobicity product (0.524), charge attraction (0.505), and a supervised combined logistic regression model (0.653; leave-one-out cross-validation). Wilcoxon paired tests confirm the difference is highly significant (\(p < 0.0001\)), with TOURIST winning in 102 of 145 complexes. The method performs substantially better on complexes exhibiting lock-and-key geometry (mean normal dot product \(< -0.30\): AUC \(= 0.765\), \(n = 114\)) than on non-lock-and-key complexes (AUC \(= 0.562\), \(n = 31\)). Bound and unbound performance are nearly identical (0.722 vs 0.733), indicating robustness to conformational change upon binding. We characterize the scope of TOURIST through a pose discrimination experiment: the descriptor does not distinguish near-native from non-native docking decoys (mean AUC \(= 0.474\) on ZDOCK decoy sets), confirming that it captures pairwise interface complementarity rather than global pose quality. This defines TOURIST as a contact prediction method for unbound structures rather than a docking scoring function. Nevertheless, the unbound contact prediction performance demonstrates that geometric shape complementarity, encoded as octant-based SH descriptors, captures structural information beyond surface exposure and physicochemical properties alone. A companion web application provides interactive visualization of lock-and-key complementarity and geometric docking of unbound protein structures.

Preprints.org
Protein Structure and Dynamics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon