Enki

Enki™ is a proprietary generative AI foundation model developed by Variational AI, purpose-built for small molecule drug discovery. Trained directly on molecular structures and properties from hundreds of millions of curated experimental and computational samples spanning 700+ drug targets, Enki™ generates novel, drug-like compounds that are inherently optimized to meet a defined Target Product Profile (TPP), including potency, selectivity, ADMET properties, and synthetic feasibility.

Enki™ is designed for R&D teams working across early drug discovery, from initial hit identification through to lead optimization, and is particularly well-suited for teams working on novel or difficult targets where experimental data may be sparse, noisy, or entirely absent.

How Enki™ Works

1. Define the preclinical Target Product Profile (TPP): Users specify on-targets and off-targets, as well as the desired physico-chemical properties of the molecules, providing Enki™ with the parameters needed to guide compound generation.
2. Enki™ generates compounds: The model produces novel and structurally diverse small molecules that meet the defined TPP, navigating uncharted chemical space to surface candidates unlikely to emerge from traditional screening methods.
3. Make your selection: Chemists review and select compounds for synthesis and testing. Enki™ can also perform hyper-efficient lead optimization constrained to a defined scaffold, enhancing key properties without altering the core chemical structure.

Key Capabilities and Differentiators

• Unprecedented novelty: Enki™ escapes the constraints of legacy screening libraries by exploring uncharted regions of chemical space, discovering structures that traditional methods are unlikely to produce.
• Drug-like and synthesizable outputs: The platform prioritizes synthetic accessibility and property balance, delivering real, actionable candidates rather than theoretical structures.
• Faster hit generation: Enki™ can deliver 100+ high-quality hits in weeks rather than months, significantly accelerating early discovery timelines.
• Minimal data requirements: The model is effective even when experimental data is sparse, noisy, or non-existent, making it ideal for novel or difficult targets.
• Flexible, human-in-the-loop workflow: Enki™ integrates into existing discovery processes as a design-on-demand partner, allowing chemists to guide, review, and iterate on generated outputs.
• Scaffold-constrained lead optimization: The platform supports hyper-efficient lead optimization within a defined scaffold, enabling property enhancement while preserving the core chemical structure.

Supported Use Cases

• CNS-penetrant ATR inhibitors: In collaboration with Rakovina Therapeutics, Variational AI has used Enki™ to advance a hit series of ATR inhibitors, generating compounds that balance potency, selectivity, stability, and early CNS exposure.
• Selective dual EGFR/FGFR1 inhibitors: Variational AI partnered with Life Chemicals to discover de novo generated selective dual EGFR/FGFR1 inhibitors using Enki™, building on a prior successful collaboration targeting the SARS-CoV-2 main protease.

Enki™ supports R&D teams throughout the early drug discovery process, from identifying novel hits to optimizing leads, enabling efficient exploration and refinement of small molecules at every key stage prior to preclinical development.