BindCraft reimagines protein binder discovery through structural prediction

Excess carbohydrates in the diet are converted into triglycerides, which involves the synthesis of fatty acids from acetyl-CoA in a process known as lipogenesis, and takes place in the endoplasmic reticulum. In animals and fungi, a single multi-functional protein handles most of these processes, while bacteria utilize multiple separate enzymes. Some types of unsaturated fatty acids cannot be synthesized in mammalian cells, and so must be consumed as part of the diet, such as omega-3. Acetyl-CoA is also involved in the mevalonate pathway, responsible for producing a wide range of isoprenoids, which include important lipids such as cholesterol and steroid hormones.

The main protein or peptide in Alzheimer's disease patients is something called beta-amyloid. It has a 42 amino acid chain, and accumulates in the plaque between nerve cells in Alzheimer's disease patients. Beta-amyloid has been found to epimerize. Certain amino acids epimerize in the beta-amyloid at specific locations, and it turns out those amino acids are mainly aspartic acid and serine. There was no good way to analyze for such aberrations in beta-amyloid until now, and we have found a number of techniques that not only detect it, but can point out exactly where, within the beta-amyloid chain, the aberration is occurring and if there are more aberrations in Alzheimer's patients than in normal people. Image Credit:Shutterstock/Tavarius

One major example is protein structure. Methods such as AlphaFold and Boltz-2 have transformed how we generate 3D structural data. Accurate protein structure prediction has been a decades-old challenge, and while there is still work to do on underrepresented protein classes, these models have opened an extraordinary pathway. We have also seen progress in multi-output models. Just over a decade ago (circa 2013), toxicology and property prediction models often focused on one endpoint at a time. Now, multi-output models, especially from methods such as graph neural networks (GNNs) and other deep learning methods, can predict multiple endpoints simultaneously. This allows for transfer learning between related properties. That creates opportunities across ADMET (absorption, distribution, metabolism, excretion, and toxicity) prediction and toxicology. Another exciting area is machine learning potentials. Molecular simulations traditionally rely on fixed mathematical functions to describe molecular interactions, but neural potentials can now replace or augment those functions. They can, in many cases, provide highly accurate simulations and structures at lower computational cost than quantum chemistry. However, its not just down to model architectures. The biggest blind spot behind all of these advances remains careful data curation, annotation, and collection. Open databases for example ChEMBL, PubChem and Chemspider together with data initiatives such as OpenADEMT and OpenBind to name but a few are making notable progress here. We still see some data for example on formulation chemistry often receiving less attention. Its is not just about data scale, but quality that is critical here to. Accessible high quality data sets are the fuel of AI methods. Architecture matters, but high-quality, accessible and abundant data is also absolutely critical.

What we're trying to do here at Benevolent is develop systems and processes that allow us to tap into the huge amount of public information and some propriety databases that we have paid to access and through that, generate hypotheses which our scientists can then triage. We've got over ten times the amount of information in our database at present (and it’s growing) than we believe that companies such as Watson have. Bringing that and filtering the signals from the ‘noise’ to our experienced drug discovery chemists and biologists, means, firstly, we can reduce the amount of time we take to triage new hypotheses, and secondly, we can develop more predictive chemical tools to allow us to make better molecules and also to reposition existing molecules for new indication.

The simplest motifs are composed of multiple secondary structure units. These simple motifs can contain α-helices and β-sheets that are layered adjacent to one another either in the same direction (parallel) or in the opposite direction (anti-parallel). The simplest motif is the formation of a “loop”, known as a β-turn if it is short, while an unstructured connection is termed a “coiled region”.