Dock

• 1. Introduction

  ›

  DOCK is a molecular modelling program with a singular claim to history: it was the first docking software ever built. Created in the 1980s by the research group of Irwin "Tack" Kuntz at the University of California San Francisco, it arrived before anyone had a standard approach to predicting how small molecules bind to biological targets. The central question it set out to answer was deceptively simple: given a protein pocket and a candidate drug molecule, where does that molecule land, and how tightly does it hold? Answering that question turned out to require building algorithms that nobody had assembled before. What did those algorithms actually look like, and how has the program continued to evolve in the decades since its creation?

• 2. Geometric Foundations

  ›

  DOCK uses geometric algorithms at its core, relying on shape as the primary language for describing a molecular interaction. In the method known as rigid docking, the program places spheres inside the binding pocket of a target protein, then performs what is called bipartite matching between those spheres and the incoming small molecule. This approach treats both the pocket and the ligand as fixed shapes and asks how well they fit together. Bipartite matching is a graph-theory technique that pairs two sets of objects optimally, and applying it to three-dimensional molecular geometry was a significant conceptual step for the field. This sphere-based shape matching is present across all versions of DOCK, making it the most durable single idea in the program's history.

• 3. Handling Flexible Molecules

  ›

  Rigid docking captures a useful first approximation, but real drug molecules are not rigid. DOCK addressed this by introducing methods that account for ligand flexibility. One is called "anchor and grow," an algorithm introduced in versions 4 through 6 of the program. The approach works by identifying a rigid substructure of the molecule, anchoring that piece in the pocket, and then incrementally building out the flexible portions. A separate method, hierarchical docking of databases, was available in versions 3.5 through 3.7, offering a way to screen large collections of compounds by processing them in stages rather than all at once. Together these two strategies gave researchers tools for handling the conformational complexity that rigid matching alone could not capture.

• 4. AMBER Score and Receptor Flexibility

  ›

  David A. Case's group took DOCK a step further by implementing a molecular dynamics engine into version 6. This addition brought a scoring function called AMBER score into the program's calculations. Where earlier scoring relied on static snapshots of the bound complex, AMBER score treats the receptor as something that can move. By sampling energetic ensembles rather than a single frozen conformation, the program can rank candidate molecules by their average binding energy across many possible receptor configurations. This is the kind of flexibility that matters most when a protein pocket breathes or shifts around a ligand. The result is a more physically realistic picture of binding affinity than any single-pose evaluation could provide.

• 5. Two Living Versions

  ›

  DOCK 6 and DOCK 3 are both actively developed today, representing two distinct lines of the original program. Codevelopers Brian K. Shoichet, David A. Case, and Robert C. Rizzo work alongside Kuntz's legacy in maintaining the software. The existence of two parallel versions reflects the different needs that have emerged in computational chemistry: one line has grown richer in physics-based scoring and flexibility modeling, while the other retains characteristics suited to large-scale virtual screening. Shoichet's involvement in particular connects the program to decades of work on structure-based drug discovery at UCSF, a lineage that runs directly back to Kuntz's original group in the 1980s.