Revolutionizing Drug Discovery: An Interview with Receptor.AI

1. Can you explain how Receptor.AI's technology works and how it differs from traditional methods of drug discovery?

Traditional drug discovery techniques are based on a combination of high-throughput screening in the lab with so-called physics-based computational methods. These techniques assess the strength of binding between the compound and the target protein using the basic laws of physics under a large amount of empirical assumptions. There are two fundamental problems with such techniques: they are inherently slow, which limits the size of the chemical libraries for virtual screening, and they only provide affinity estimates with no information about ADME, toxicity and real-world biological activity.

In contrast, Receptor.AI utilises Artificial Intelligence and Deep Learning to build data-driven models, which are able to predict not only actual biological activity but also ADME endpoints and toxicity. These models are able to screen enormous chemical spaces with tens of billions of compounds in a single day.

Moreover, in contrast to traditional techniques, our AI learns from experimental feedback at each iteration and improves itself as far as we go based on the result of experimental validation of predicted hit candidates.

2. Can you provide some examples of how Receptor.AI's technology has been used by pharmaceutical and biotech companies?

We currently have more than 10 commercial projects in the pipeline from academic institutions, small and medium pharma companies and two large CROs. The tasks range from selective anti-cancer kinase inhibitors to agonists of taste receptors for fighting obesity and the blockers of ionic channels involved in hypertension.

3. How does Receptor.AI's technology help to speed up the drug discovery and development process?

We provide optimised pipelines for all major stages of preclinical drug discovery, namely, hit discovery, lead discovery and lead optimisation, which are available as either the SaaS platform or custom services at very competitive prices.

Obtained candidate compounds are guaranteed to be cheap to synthesise or order from the vendors. During the virtual screening, we can apply up to 100 different filters on various ADMET endpoints, drug-likeness metrics and phys-chem parameters. This ensures that the resulting compounds are exactly what the customer needs, so no resources are wasted on laboratory validation of unsuitable molecules.

This allows us to dramatically reduce the time and cost of pre-clinical drug discovery stages and increase the success rate to 90%, which is way better than traditional techniques.

4. Can you discuss any specific success stories or case studies of how Receptor.AI's technology has helped a pharmaceutical or biotech company advance their drug development program?

Unfortunately, we are limited by NDAs in most cases, but we have two publicly available case studies that we used as a proof-of-the-concept for our first MVP solution. Our AI platform has designed hit compounds in fully automated mode for two well-known cancer-related drug targets: BRD4 and SIRT1 proteins. For BRD4, we have got 17 compounds with good affinity profiles and have found several active compounds from the previously unknown classes. For SIRT1, we have discovered a nanomolar hit compound with lead-like characteristics from the first shot.

5. How does Receptor.AI's technology help to improve the accuracy and efficiency of identifying new drug targets?

To identify novel targets, we leverage our proprietary knowledge graph to analyse and interpret the complexity of biological systems through the identification of functional relationships between diseases, proteins, genes, and pathways. This information is then integrated with the expertise of our team of bioinformaticians and the AI-augmented omics atlas to generate novel theoretical hypotheses. These hypotheses subsequently undergo experimental validation through collaboration with our partners. However, it is essential to note that these predictions are limited to identifying disease-associated genetic mutations in classical genetic disorders such as thalassemia and various forms of cancer. Any claims of broader detection capabilities based solely on omics data should be viewed with scepticism.

6. Can you discuss the company's partnerships or collaborations with other pharmaceutical and biotech companies or research institutions?

Receptor.ai has established a strong partnership and strategic alliance to provide an end-to-end drug discovery and development pipeline from early discovery to clinical trials. To complete any specific project, we have multiple partners related to compound synthesis, biological validation, and clinical trial automation. In addition, we collaborate with a number of academic institutions from Europe, the USA and Asia to provide free access to our SaaS platform for drug design and help them get scientific grants.?

7. Can you speak to any regulatory or ethical considerations surrounding the use of AI in drug discovery and development?

Fortunately, our technologies stay apart from the most debated ethical topics of using AI in healthcare, which are related to using patient data and prescribing treatment based on AI predictions. We work on the preclinical stage using very little data, subject to ethical regulations. All such data is public and strictly anonymised.

However, one sensitive ethical question, which our technology helps to solve to some extent, is animal testing. We still can’t develop drugs without animal tests, but such testing could be substantially reduced using our virtual screening technologies. A large part of the cruellest ADME and toxicity animal testing could be substituted by our AI models.

8. Can you discuss any current or planned developments or expansion plans for Receptor.AI?

We have ambitious plans for expansion for the coming few years. First, we are going to provide complete end-to-end AI solutions for biotech and pharmaceutical companies, which will cover the whole preclinical drug discovery pipeline. Second, we plan to establish our own wet lab, which will give us precise and complete in-house experimental feedback for improving our AI models and will allow us to collect unique proprietary data for model training and testing, which will set us apart from competitors. Third, we plan to expand to the Asian market and establish strategic collaborations with several big CROs and biological data collection companies. Finally, we have launched several joint venture end-to-end drug discovery programs with our partners to bring them to clinical trials for up to 2 years with subsequent drug asset out-licensing to big pharma.??

9. How do you see AI and machine learning impacting the pharmaceutical industry in the future?

We strongly believe that AI is the future of the pharmaceutical industry, and its role will rapidly increase in the coming years. We are heading towards fully automated drug discovery pipelines, where AI not only facilitates individual stages of the process but also guides the whole project, including the planning of experiments and their interpretation. At one point, we expect to see a breakthrough when the amount of collected data and the advances in AI architectures will result in qualitative breaking changes in the speed and efficiency of drug discovery. The success rate should eventually jump from the current 4% to 50+%.

10. Can you discuss any challenges or obstacles that Receptor.AI has faced and how the company has overcome them?

Our company originates from Ukraine, and the war initiated by Russia put a significant challenge on our operation. However, we managed to relocate the core team to Central Europe and arrange the working processes in such a way that in just a few months, we were able to continue our work as a UK legal entity. We have significantly increased our productivity and hit milestones faster than originally planned. Tough times make you move faster, adapt better, and get stronger.