Computational Drug Discovery

The physics of
water, made precise.

Euphemia combines rigorous molecular simulation with machine learning to map solvation networks that conventional docking ignores — turning water from noise into signal.

Partner with usOur science →
~10³
Water sites mapped
GCMC
Grand canonical engine
FEP+
Rigorous ΔΔG
The science

Solvation is the unsolved problem.

Every drug binds in water. The binding-site water network dictates whether a molecule displaces a thermodynamically costly water molecule, inherits a favourable one, or misses the pocket entirely. Docking scores ignore this. Euphemia does not.

Physics
Grand Canonical Monte Carlo
We use GCMC/MD to equilibrate explicit water networks across binding sites at thermodynamic resolution — finding every occupied, displáceable, and structural water position.
Machine Learning
ML-accelerated scoring
Solvation-aware descriptors trained on free energy data let us screen chemical space at scale without sacrificing physical accuracy — ML guided by physics, not replacing it.
Truth
Rigorous ΔΔG validation
Free Energy Perturbation benchmarks against crystallographic water positions and experimental affinities. Predictions are only as trustworthy as their calibration.
Our approach

Physics as the foundation.
ML as the accelerant.

The field has bifurcated into two camps: empirical ML models that train on affinity data without physical understanding, and academic simulation workflows too slow for drug discovery timescales.

Euphemia sits at neither pole. Physics sets the boundary conditions; machine learning navigates within them. The result is predictions that generalise to novel chemotypes — the cases where data-only models fail.

01
Target hydration mapping
GCMC/MD across the binding site, producing a thermodynamic water map with occupancy and ΔG per site.
02
Solvation-aware virtual screening
ML scoring functions informed by water displacement thermodynamics — not just shape complementarity.
03
FEP lead optimisation
Rigorous relative binding free energy calculations for shortlisted compounds, prioritised by experimental tractability.

Where Euphemia differs

Shape + electrostatic scoring
vs. Docking
Water displacement ΔG as primary binding signal
Interpolation within training distribution
vs. Pure ML
Physics-grounded generalisation to novel scaffolds
Publication timescales, no decision pipeline
vs. Academic MD
Campaign-ready throughput with rigour maintained
General-purpose; water networks as add-on
vs. Schrödinger / FEP+
Water network analysis as the core differentiator
Work with us

Serious about solvation?

We partner with biotech and pharma teams where water network modelling is a bottleneck — early target validation, hit-to-lead, or FEP campaign design. If you have a target and want to understand its hydration, let's talk.

info@euphemia.ai