Designing Antibodies
Why we invested in Nabla Bio
Getting even one drug from the research lab to the clinic is a multi-step process that takes an average of $2.6B and 12 years — that is if the drug has not failed before reaching full approval, a milestone less than 0.02% of potential drugs reach. To put it plainly, it is near impossible to get a new drug approved. One of the biggest conundrums surrounding biotech today is that we understand the science and mechanisms of disease better than ever before and yet we are still using outdated methods for developing new therapeutics.
The drug development process is broken into four main phases: drug discovery, preclinical studies, clinical trials, and regulatory review. The major bottleneck is in clinical trials — you just can’t speed up time when tracking patient responses. However, by increasing the chances a drug will be successful in clinical trials, we can make the drug development process more efficient — saving years and billions in what would otherwise be sunk costs. To do so, the challenge lies in drug discovery — designing the right compound and choosing the right asset to advance to trials. This is not so straightforward. Drug design involves finding the right biological target (a cellular structure involved in controlling disease biology) and then finding a corresponding lead compound (small molecule or protein that acts on the target to have a desired effect). Given that humans have around 20,000 genes and can express over 200,000 proteins, finding the right combination of target and lead is no small feat.
Luckily, computational biology has evolved to the point where we can let machine learning algorithms do a good chunk of the work — we no longer need to take our best guess and screen hundreds of potential drug targets and candidates by hand to find an optimal target or lead. Algorithms can screen far more of the genome, proteome, and even epigenome in a fraction of the time that experiments need to find druggable targets and corresponding candidates. Companies like Recursion, LabGenius, Atomwise, and Cantos partner companies Phenomic and Clover Therapeutics are all applying AI to drug discovery in this way. However, the problem is not fully solved there. Once a lead candidate is identified, it must be optimized to ensure sufficient target binding, low toxicity, stability, etc — a set of features collectively referred to as developability parameters. When considering protein drugs alone, lack of sufficient optimization accounts for 30% of clinical-stage failures. That is $450M lost per drug solely due to lack of optimization.
Nabla Bio is fixing this. Spun out of Co-Founder & CEO Surge Biswas’ PhD research at Harvard, Nabla is building Autoverse, a protein design platform that enables protein optimization using a small number of data points (low-N). Essentially, Nabla’s Autoverse can design proteins from scratch with minimal prior knowledge — solving the decades long biotech question of how do proteins go from sequence to structure and providing useful insights for drug optimization when desired properties are known but structure is not.
Such a platform can be used for a multitude of applications within biotech, synbio, agriculture, and beyond. Nabla’s core focus is on accelerating drug development, specifically for antibody-based therapeutics. Antibody drugs are more dynamic than other classes of therapeutics due to their size and programmability. There are several blockbuster monoclonal antibody drugs (ex: Humira) on the market today that have proven the success of these protein drugs.
However, optimizing developability properties of antibodies is a time-consuming, difficult process that often results in clinical failures. Moreover, promising next-generation antibody therapeutics such as bispecifics and nanobodies require increasingly complex engineering that has made them difficult to advance with current technologies. Nabla makes engineering developability and next-gen antibodies simple with their AI + wet lab approach. The team has proven they can design the most developable antibodies with at least 35x higher efficiency than industry standards in a fraction of the time.
It’s clear that Nabla’s technology is special, but that’s not what attracted us to the company in the first place. Cantos was created to support world-class scientists and entrepreneurs in their quest to build the near frontier, and Surge and his Co-Founder Frances are definitionally that. Together with advisors Mohammed AlQuraishi and George Church, Surge is the first to apply natural language processing (NLP) algorithms to protein design. He has a deep understanding of the relationships between protein sequence, structure, and function and has used machine learning to uncover the missing links between sequence and structure that have long been a focal point of research in the scientific community. Combined with Frances’ experience in venture creation and protein discovery at Flagship Pioneering and Novartis, this is the team you want designing your protein therapeutics.
Nabla is bringing us one step closer towards the ultimate goal of deliberately designing drugs, not just discovering them — improving therapeutic power and significantly increasing the probability that a given drug candidate will make it to approval. We believe Nabla and its computational biology peers will save lives and ultimately reduce the cost of treatment. With their combined AI + wetlab approach, Nabla can design better biologics faster and cheaper than previously thought possible. We are proud to partner with Surge, Frances, Zetta Venture Partners, and Khosla Ventures as well as existing investors Fifty Years and Boom Capital in Nabla’s $11M seed round.
