Binding
Affinity
balancing speed, accuracy, simplicity
Koffee™ is our computational drug discovery technology for classical computers — delivering results faster than it takes to brew a cup of coffee, and with no quantum hardware required yet.
Koffee™ Binding Affinity ranks small molecules by binding strength — balancing speed, accuracy, and simplicity for modern drug discovery.
Traditional free energy methods like Free Energy Perturbation are accurate but slow and expensive. Docking scoring functions and machine learning models are fast, but often unreliable.
Koffee™ Binding Affinity bridges this gap with fast, automated predictions in under a minute on a standard CPU — with no training or target-specific tuning required. Built on a modern, physics-based force field with support for polarization and precise hydrogen and halogen bonding, Koffee™ Binding Affinity delivers robust performance across diverse targets for analog or congeneric ligand series.
We are currently working on the next major version of Koffee™ Binding Affinity. Sign up now to be informed when it is available and be among the first to experience its new features and improvements.
Performance on Public Datasets
Koffee™ Binding Affinity achieves robust correlation across diverse targets and ligand chemotypes.
In particular, it performs well for analog/congeneric series — in several cases demonstrating comparable accuracy to much slower methods.
Koffee™ Binding Affinity is orders of magnitude faster than FEP methods. The average run time per complex is ~30 seconds on our extended test set on an 8-core CPU machine.
Prospective Performance
In collaboration with a pharmaceutical partner, Koffee™ Binding Affinity achieved a Spearman correlation coefficient (ρ) of 0.6 on previously characterized (unblinded) compounds and 0.77 on a blinded test set.
While validation against experimental data remains important for confirming protein structures and binding modes, Koffee™ Binding Affinity does not require target-specific fine-tuning.
These results highlight Koffee™ Binding Affinity's potential for prospective use in drug discovery pipelines.
Use Cases
Rank Congeneric Ligand Series
Koffee™ Binding Affinity provides rapid, physics-based ranking of structural variations within congeneric ligand series, identifying the most promising analogs during Design-Make-Test-Analyze cycles.
Prioritize Ligands in Generative Chemistry
Modern generative design tools can propose a vast number of molecular modifications — but distinguishing meaningful improvements from random changes remains difficult.
Complement FEP Workflows
Use Koffee™ Binding Affinity as a fast, physics-based filter to focus more computationally heavy Free Energy Perturbation (FEP) calculations on the most promising candidates.
Integrate with Existing Pipelines
Delivered as a Docker container with no external dependencies, Koffee™ Binding Affinity ensures full control over sensitive data while integrating easily into computational pipelines.