DLMS
In DevelopmentA platform for learning and modeling systems that evolve over time. Temporal dynamics as a first-class primitive: predict trajectories, learn interaction potentials, and capture time-dependent structure from data.
The first engine built on DLMS is MDE, Vareon's Molecular Dynamics Engine.
Molecular Dynamics Engine
Structure prediction tells you where atoms sit. MDE tells you how they move, how they interact, and what happens next. It predicts molecular dynamics at timescales and costs that conventional simulation cannot reach.
Predict how molecular systems evolve over time, not just equilibrium positions. Dynamics, transitions, and kinetic pathways are first-class outputs.
Build interaction models directly from your data. No manual parameterization required.
Generated trajectories conserve energy, respect symmetries, and remain stable over long simulation horizons.
Answers how atoms got where they are, what happens next, and how the system responds to perturbation.
Post-AlphaFold
Static structure prediction changed the field. The next problem is understanding how molecular systems behave over time.
AlphaFold and its successors predict static protein structures with high accuracy. Folding pathways, conformational changes, binding kinetics, and allosteric regulation all depend on dynamics that static snapshots cannot capture.
Binding affinity depends on how a ligand approaches and sits in a pocket over time, not just the lowest-energy pose. Residence time, induced fit, and off-target transitions are inherently dynamic properties that require trajectory-level modeling.
Conventional simulation can model these processes but at enormous computational cost. Biologically relevant timescales remain out of reach for most systems. MDE compresses that cost.
Raw Capability
MDE learns to simulate molecular systems by training on trajectory data. Given initial conditions, it predicts how a system evolves: conformational transitions, binding events, folding pathways, and response to perturbation.
Physically consistent trajectories at a fraction of the cost of conventional simulation, extending accessible timescales by orders of magnitude.
Target applications
DLMS platform foundation
MDE is built on the DLMS platform layer, sharing core infrastructure across engines.
Platform
A shared computational framework for any system where behavior over time is the modeling target. MDE is the first engine. The platform extends to any domain where dynamics carry the signal.
A shared computational framework for systems that change over time. The same learning primitives serve molecular, biological, and engineered systems.
MDE is the first engine on the DLMS platform. Each engine brings its own domain science on top of shared infrastructure.
Every trajectory, learned potential, and modeling decision is traceable. DLMS supports auditable workflows from research through production.
DLMS and MDE are in active development. If your work involves molecular dynamics or long-timescale simulation, we want to hear from you.