We matintain a library of specialised solutions, and can build bespoke solutions at speed.

The following sections outline our development pipeline and technical expertise. 

We mine ultra-vast chemical spaces for novel candidate entities, using high-performance computing to systematically screen all new chemicals as soon as they are discovered. We evaluate synthetic accessibility early in our validation pipeline.

Our custom crawlers trawl the deep web, collating overlooked data from a source roughly 400-times larger than the surface web.  

Alongside this, we use scaffold-hopping to find previously unknown analogues of key reference chemicals. Our custom-coded deep-learning neural networks have been trained to identify activity-enchacing structural modifications across types and scales.

We prioritise areas free of existing intellectual property claims, whilst finding important new analogues in well-established domains. 

Our systems run continuously, and were built to ensure data sovereignty and security.

For evaluation in silico, we stream encrypted data analysis workloads from quantum-secure servers into our state-of-the-art mathematical models.

Biological characterisation in vitro uses innovative high-throughput systems developed for the rapid iteration of chemicals across target tissues and cell-types. Alongside this, we run proprietary high content systems for specific compound classes, such as ice-recrystallisation inhibitors.

We make empirical measurements of key biophysical characteristics, because many of these are affected in a dose-dependent manner by cryoprotectants. We consistently find that values for parameters such as permeability, diffusivity, and viscosity – the foundations upon which biophysical models are built – diverge substantially from those reported in the published literature. Using the correct values has far-reaching implications: protocols developed by others using inaccurate numbers simply cannot succeed.

We apply these measurements to develop system-specific models which address mass-transport, network thermodynamics and diffusion. We juxtapose models of diffusion: concentration-dependent, non-Fickian, and as spaning heterogenous media with moving boundaries. We use molecular dynamic models and, with our academic collaborators, a range of spectroscopy methods.

By using a  wide-range of methods we reach conclusions that are reliable and robust.

As candidate chemicals emerge, we characterise their interactions with each other. We use optimization techniques to delineate non-linear interactions, characterising relationships with key outcomes to ensure our robust definition of reproducible global minima. Gaussian processes and the extraction of random Fourier features allow us to understand the parameter spaces within which we operate.