An emerging business model in drug development
Software-first, generative biodesign
Discovery vs design
Over the past few months I have spent some time mapping out the landscape of companies at the intersection of biology and AI. Originally, I shared the idea that AI tools are best suited for the design phase of drug development, as opposed to discovery. I detailed the core of this hypothesis in my post The evolution of ML in early-stage drug development:
Over the past few years I believe there has been a shift from an emphasis on ML-driven discovery to ML-driven design. Initially, ML was seen as a potential disruptor for scientific discovery, promising to unveil new mechanisms and insights into disease biology. However, the field has undergone a recalibration of expectations, with ML now finding a narrower, albeit more applicable role in design-oriented processes. This represents a transition from a 'first-in-class' to a 'best-in-class' approach, where ML contributes to accelerating design iterations for established targets and known biology.
I go on to discuss how the design cycle is better suited to experimental lab result feedback, and sketch out how current market dynamics may influence how this hypothesis is tested in the real world. Recently I have been considering how design-focused businesses will approach this industry dominated by therapeutic asset ownership.
This article explores how an emerging cohort of software-first, generative biodesign companies could approach the drug development value chain. But first, it's important to delineate the major market verticals and look at the distribution of existing strategies.
Assets rule everything around me
Historically, the biotech sector has been a unimodal distribution of business strategies. The majority of value has been created through the ownership of therapeutic assets. This happens through internal development or strategic partnerships/acquisitions. Revenue generated from these sales supports a broad network of activities, including internal product development, partnerships, acquisitions, outsourced scientific services, and tools to facilitate these processes.
Typically, a large amount of capital is invested in developing internal assets. Alternative strategies to break into drug development remain bleak. A perennial question persists: could a new strategy disrupt the value chain? Given the significant impact computation has had on other industries, many believe that it will similarly revolutionize drug development.
The largest bet to date has been placed on computational technology supporting internal asset development. The hypothesis is compelling: we can measure biology at scale, allowing us to accurately model biological processes. If this works it should lead to significant improvements in discovery efficiency. Go after the big prize.
Is this the only pathway to creating value? The current generation of AI-native drug development companies has shifted their focus to using AI tools for design rather than discovery. The generate, test, iterate cycle meshes well with experimental laboratory frameworks. Interestingly, there appears to be a trend among teams to apply generative design technologies beyond internal R&D.
Generative biodesign firms
I have noticed a trend of teams testing whether alternative approaches to value creation in biotechnology are viable. Most of these are early-stage startups (BioLM, 310.ai, Cradle, Evolutionary Scale, Profluent), save for one big player (NVIDIA).
These are teams developing robust generative software products and testing them across various market verticals. The framework looks like this: a software that serves generative models (mainly small molecule & protein) and internal efforts to improve these models.
Of course there are many new companies founded upon generative design tools (Isomorphic Labs, Absci, Xaira) but these groups appear to be pursuing traditional, asset-first strategies.
The magic of software lies in scalability. So you can imagine scenarios in which such products are A) utilized to develop internal assets, B) leveraged as a competitive advantage in partnership deals, C) offered as dedicated services for outsourced development, or D) marketed to a base of 'consumers'—ie, researchers.
Alternative paths
While priors suggest that the value hierarchy will not undergo significant changes, it's worth exploring alternative strategies beyond vanilla internal asset development.
One new idea is that these generative design products could be marketed directly to consumers, resembling a pure SaaS model. This is how a few of the software products are being marketed right now, for example NVIDIA’s BioNemo. The potential benefit is the democratization of asset development, leading to a more accessible and scalable drug development industry. This could result in more, and better, therapies in a very competitive marketplace. Kind of a crazy future.
Currently, a major limitation is the relatively small consumer base compared to other SaaS markets. This further narrows considering the subset of teams who have the R&D infrastructure to participate in the physical, experimental testing feedback cycle. Also, these consumers are (skeptical) researchers, who must realize true problem-solving benefits.
Heading up the strategic chain, we enter the markets of services and partnerships. If a shift away from an asset-first approach is possible, I think there are interesting opportunities here.
Proprietary software and design expertise could enable the establishment of a significantly higher number of partnerships than traditionally possible. This could create more shots on goal, and better accommodate the power-law dynamics of drug development.
I recently discussed this idea with a colleague and a few comparables were brought up: Schrödinger and Atomwise. Schrödinger has traditionally operated a computational chemistry software business through SaaS and service models. Recently they signaled a strategic shift towards internal drug development programs. This move underscores the notion that "all roads lead to drug development."
Atomwise, a first-gen small molecule design firm, pursued a unique “many partnerships” approach, inking small deals across academia and industry. From a statistical perspective, this strategy makes good sense.
However, Atomwise's technology may not have been sufficiently mature to accelerate multiple partner programs in parallel. This underscores how design is just one aspect of drug development, and frankly, not the most significant bottleneck. By spreading resources across many modest partnerships, it may have become difficult to concentrate efforts on the most promising candidates. In an industry still centered around therapeutic assets, going all-in to create outlier outcomes is essential.
Maybe there is a mixed strategy within the spectrum of SaaS and partnership. A dual-pronged approach could involve wide distribution of the software product to support a cash-positive business, alongside a moderate number of strategic partnerships. By carefully selecting these partners and programs, focusing on strong leads, this blend of strategies could help bolster asset development. It remains to be seen if these new generative design technologies are truly impactful in accelerating time-to-conclusion of research programs.
But the data…
Critical questions revolve around data. This is the engine oil for AI-first businesses. While collecting data internally is feasible, this would again require substantial capital (in the absence of a novel, cost-effective platform). An intriguing approach is exemplified by Basecamp, which has been sampling soil worldwide to build a dataset of unique protein sequences. Another plausible model might be offering a "freemium" software product in exchange for data during the design-test feedback cycle.
Imagining these potential futures and their disruptions is intriguing, yet predicting how they might realistically unfold is challenging, given the high capital requirements and binary outcomes in drug development. The industry staunchly operates in a winner-take-all manner, with limited avenues for incrementally growing value along the way. Nevertheless, teams exploring new strategies in this space are an intriguing experiment in play with a bold vision for the future.