Why it matters
The BenevolentAI ALS Programme
ALS, also known as motor neurone disease or Lou Gehrig's disease, is a devastating condition which causes the death of neurons controlling voluntary muscles, leading to difficulty speaking, swallowing, and eventually breathing. There is no cure, and current treatments only average a three month extension of life. BenevolentAI set up a dedicated team to explore new approaches.
By approaching a poorly understood and enormously complicated disease, we knew any progress our platform could make in finding an effective treatment would become powerful proof of the potential for machine learning in drug discovery, and also of our potential to change patients’ lives.
New target validation
BenevolentAI's platform produced a ranked list of potential ALS treatments, together with biological evidence. Our team, with no previous specific expertise in ALS, were able to rapidly triage these predictions using strategies focused on pathways implicated in multiple ALS processes. The five most promising compounds were taken to the Sheffield Institute for Translational Neuroscience (SITraN), a world authority on ALS. An ALS candidate emerged from a breast cancer drug, which showed delay of symptom onset when tested in the gold standard disease model.
This was a strong proof of concept that suggested multiple routes forward; optimisation of the hit compound is underway, and our machine learning and data-led approach to target identification was developed further and reapplied to ALS. The ALS programme led directly to improved workflows and algorithms that can be applied to any disease — this is at the core of what we strive to do with our platform. This year we re-applied these evolved approaches within the ALS programme to identify further routes forward.
ALS is a hugely challenging disease, heterogeneous across cell-types, genetics and patients. Drug development requires a patient-derived phenotypic assay to validate biological hypotheses and BenevolentAI is collaborating with SITraN, one of the world leading centres for research into ALS, using the gold standard disease models as part of our programme.
Fast lead optimisation
Our platform also supports the ALS programme through lead optimisation to identify a clinical candidate. EvoChem is a tool that generates de novo compound ideas based on complex multiparameter optimisations specified by our medicinal chemists, enriching the ‘chemical palette’ available to them during the process. This and other proprietary molecular design tools reduce the number of compounds synthesised and reduce the number of experimental cycles required per programme. Usually, designing a viable drug candidate takes 4-8 years. We achieved this in 14 months on a number of BenevolentAI programmes. For ALS, we want to identify a clinical candidate in under a year.
This AI-augmented molecular design platform is already helping to design active compounds utilising multiple algorithms. We are generating results with greater and more diverse polypharmacology than our first ALS candidate, and our most recent workflows for hypothesis generation have resulted in our current lead, BEN-2349. This compound has improved CNS exposure and is around 10x more potent than previous compounds, with a profound rescue effect in ALS patient cells.
The BenevolentAI platform has identified novel drug targets and novel lead molecules that may be potential treatments for ALS.
smart patient stratification
AI's ability to access and analyse diverse datasets can also bring opportunities in greatly enhanced patient stratification. This will improve clinical trials and avoid wasting patients’ time: good drugs often fail trials because they are being given to the wrong people, a contributing factor in the 50% of drugs still failing in late stage trials.
Benevolent's platform used PRO-ACT (a large, multi-trial ALS database) to contextualise and reason on this data. This enabled us to determine progression endotypes to inform our pre-clinical and future clinical development. We are currently working with the ALS Foundation to explore further collaborations with our Translational Medicine Squad, dedicated to building and reasoning upon deep patient-level and ‘omics datasets.
Radical cross-functional collaboration
Building the right team means harnessing the right mix of broad biology, chemistry, informatics and drug discovery perspectives. Our ALS team is evenly split between wet lab and in silico development. It requires machine learning experts who can tackle the most demanding challenges, working constantly with scientists who develop and embrace new ideas.
Acting as a catalyst for scientists and their research, our platform represents a game-changing way to evolve drug discovery. We believe the best outcomes for patients will come from those who invest in applying machine learning across the full discovery and development process, from target validation to patient stratification. This needs to be complemented by curious, ambitious teams who adopt an innovation mindset and inclusive practices.
Our ALS programme shows how machine learning and advanced technology, a culture built for innovation from the start, and radical collaboration across disciplines can create hugely promising advances towards treatment in this most difficult of diseases.