Atlas unprecedented precision at all stages reduces cycle time

Target characterization
3 months

Microsecond timescale MD with adaptive sampling allowing novel pocket identification

Unsupervised adaptive sampling strategy capable of enhancing the sampling in large biosystems

Exhaustive & ultra precise identification of long distance interactions driving allosteric modulation

Hit screening & validation
3-6 months

 

Accurate binding prediction supports medchem & AI driven compounds generation

Allows handling of complex targets such as metalloenzymes, DNA, RNA, membranes, Protein Protein Interactions, etc.

Lead optimisation
12 months

We use FEP informed AI algorithms & Medicinal chemistry to drive chemical expansion

We are developing proprietary druggability, cross-reactivity and safety filters into Atlas

Core Technology

  • AMOEBA: Advanced polarizable multipole force fields
  • Tinker–HP: HPC & Quantum accelerated Molecular Dynamics
  • Advanced sampling methodologies
  • Large scale visualisation with VTX software

Biological Space

  • Improve knowledge of the target
  • Improve multi-target drug design and polypharmacology
  • Explore dark targets & novel promising areas of the genome
  • Allow targeting of allosteric modulation

Chemical Space

  • Expand chemical space and explore neglected regions
  • Rational design and screen of ultra-large chemical libraries
  • Rescue missing hits and lead compounds
  • Improved property prediction

We have developed dual traditional and blended AI/MD screening approaches

What makes the AMOEBA force fields superior

The AMOEBA polarizable force field uses atomic multipole moment through quadrupoles and induced dipole polarization to provide electrostatic potentials accurate to 0.1 to 0.01%, compared to 1% to 10% errors for simple partial charge only models.

High-order atomic multipoles provide anisotropy critical to accurate description of hydrogen bonding. For example, isotropic models overestimate the strength of linear vs. bent hydrogen bonds at carbonyls by about 1.5 kcal/mol.

Polarization effects are very important for modelling of simple ions, metals and highly charged species like nucleic acids and membrane head groups. For example, roughly half the attraction in π-cation interactions is due to polarization, and this can amount to several kcal/mol.

Inclusion of water polarization is mandatory to model the thermodynamics of desolvation. Traditional fixed charge bulk water models such as TIP3P and TIP4P do not respond to presence in protein binding pockets, leading to significant errors in predicted absolute ligand binding free energies.

Tinker-HP brings MD with polarizable force fields at industrial scale

Tinker-HP performance and single node scalability with the AMOEBA force field

  • Linear scaling from small molecule to large viruses
  • Delivers hundred of nanoseconds of simulations per day on GPUs and microseconds per week with QM like accuracy
  • Robust and accurate algorithmics designed for polarizable models and Neural Networks
  • Optimised implementation of AMOEBA model
  • Platform independant & validated on various architectures from lab clusters to supercomputers
  • Fully benefit from the massive increase in computational power of latest GPU architectures

“Tinker-HP : Accelerating Molecular Dynamics Simulations of Large Complex Systems with Advanced Point Dipole Polarizable Force Fields using GPUs and Multi-GPUs systems”, J. Chem. Theory Comput., 17, 2034-2053 (2021)

Gaia, our in-house supercomputer
enables delivery of compounds at a massive scale

  • Target of 15 DGX A-100 – latest and fastest GPUs to date – for up to 75 pflops of computing power.
  • Linear scaling to handle massive biological bodies such as viruses and map allosteric pockets on any target
  • Up to 100 ns of MD simulation per day with polarizable force fields, enabling their routine use
  • Allows us to reach pharma-compatible throughput

We invest in quantum computing to maintain our technological edge

We use hybrid Quantum-HPC computation platform to improve simulation of complex targets

We develop novel algorithms to accelerate the discovery of novel drugs

We are open to R&D partnerships to advance the field