THe FCI company believes automated drug discovery relies on fast and accurate tools for learning chemical behaviour.

In-sillico screening is used to find “hits”, possible drug candidates. Current automated drug discovery tools can produce vast numbers of leads, but the majority of spend in drug development is on failures during live trials, with only 1 out of 8 drugs that start live trials resulting in a licence application[1].

Existing computational chemistry tools can provide excellent agreement with experiment when interpreted by experts, but have narrow domains of applicability[2][3]. Verification of the accuracy of results is only gained by using a basket of methods and applying expert judgement as to if the results tend to agree up to the intrinsic error, or if excess variance indicates a mismatch of theory; a approach that cannot be taken by a automated discovery system.

CAMBRIA provides access to ground truth simulations at lower cost and with higher accuracy than wet-lab experiments, in a package ideally suited to integration into automated learning and discovery systems and AI. Using bespoke hardware to accelerate full configuration interaction Monte-Carlo simulations, we can provide guaranteed chemical accuracy access to the full energy spectra for any molecule, with a predicted 100x speedup when compared to the latest methods running on a modern rack-scale HPC cluster.

We hope this will enable the simulation of binding sites, transition metal complexes, and other critical sections that have been inacessible to current chemical simulation tools.

We are systematically improvable, and by directly solving the electronic structure problem we can accommodate any level of theory - including first quantisation, beyond Born-Oppenheimer, and highly multi-reference systems.

[1] Innovation in the pharmaceutical industry: New estimates of R&D costs; 10.1016/j.jhealeco.2016.01.012

[2] Replacing hybrid density functional theory: motivation and recent advances; 10.1039/d0cs01074j

[3] Comparison of Hartree-Fock, Density Functional, Møller-Plesset Perturbation, Coupled Cluster, and Configuration Interaction Methods for the Migratory Insertion of Nitric Oxide into a Cobalt-Carbon Bond; 10.1021/jp961558p