AI Accelerates Biotech Drug Discovery Through Workflow Systematization - Episode Hero Image

AI Accelerates Biotech Drug Discovery Through Workflow Systematization

No Priors: Artificial Intelligence | Technology | Startups · · Listen to Original Episode →
Original Title:

TL;DR

  • AI agents can accelerate drug discovery by automating complex tasks, enabling scientists to ask questions that previously took weeks or months in mere hours, thereby unlocking organizational memory and making scientific data reusable.
  • The high cost and failure rate of drug development, estimated at over $2 billion and 7-10 years per drug, necessitates AI-driven improvements in speed and cost to make the process more efficient and attractive to investors.
  • Benchling AI's simulation tools and recommendation engines empower wet lab scientists to use complex models directly within their workflows, linking results to existing data and suggesting optimal next experiments.
  • The biotech industry's artisanal nature, with bespoke workflows and a focus on short-term survival until acquisition, creates a significant opportunity for AI to systematize processes and enable scalable, durable innovation.
  • AI's potential to improve drug discovery lies not just in clinical trials but in enhancing earlier stages, such as molecule generation and manufacturing process optimization, by making each step faster, cheaper, and more effective.
  • The "returns to intelligence" in science are substantial, as demonstrated by the success of GLP-1s and Keytruda, highlighting how novel insights and conviction can unlock transformative medicines from existing scientific knowledge.
  • Large pharma companies possess a significant advantage in AI adoption due to their unparalleled data generation capabilities, enabling them to train proprietary models at a scale that most startups cannot match.

Deep Dive

The biotechnology industry faces a critical inflection point, driven by the prohibitive cost and lengthy timelines of drug development, which AI is poised to fundamentally reshape by increasing efficiency and accelerating innovation. Benchling, as a central platform for biotech R&D, is strategically positioned to leverage AI agents and simulation tools to empower scientists, shorten development cycles, and enable the creation of more effective medicines. This transformation is not merely about technological advancement but also about a necessary recalibration of industry processes to meet a growing demand for faster, cheaper, and more accessible therapies, a shift also influenced by China's rising competitive role in the global pharmaceutical landscape.

The current biotech landscape is characterized by a significant "trough of disillusionment" following a period of intense investor exuberance. While novel therapeutic modalities like gene editing and RNA therapies have emerged, their commercial success has been slower and more costly than anticipated, leading to capital retrenchment. This has created immense pressure for the industry to become more efficient, a challenge exacerbated by the artisanal nature of much of biotech R&D, where bespoke workflows and data fragmentation hinder scalability. China's increasing speed and cost-efficiency in early-stage drug development are further compelling Western biotechs to innovate rapidly, shifting the industry's focus from novel modalities to operational excellence and speed.

AI offers a powerful solution to these systemic inefficiencies. Benchling's AI initiatives, including simulation tools and deep research agents, aim to augment scientists' capabilities by making complex computational models accessible and by unlocking institutional knowledge trapped in historical data. These tools can significantly reduce the time and cost associated with experimentation and decision-making, allowing scientists to run more experiments, identify better molecules, and compress the multi-year drug development timeline. The ultimate goal is to move from a highly bespoke, slow process to one where AI agents act as co-scientists, accelerating discovery and development, and potentially reducing the cost of bringing a drug to market from over a decade and billions of dollars to a fraction of that.

The integration of AI also necessitates a profound shift in how biotech companies operate and how data is managed. Benchling's role as a system of record is crucial, providing the structured data foundation necessary for AI models to function effectively and for scientists to trust AI-generated insights. This move toward systematization, driven by the need for efficiency and the potential of AI, contrasts with the historical "build for acquisition" mentality of many biotech startups. Furthermore, the successful adoption of AI in this complex, regulated field hinges on translating advanced AI capabilities into user-friendly, trustworthy tools that integrate seamlessly into scientists' workflows, bridging the gap between computational power and practical laboratory application.

The broader implications extend to the business models within biotech and the cultural exchange between tech and bio industries. While AI model building may commoditize, companies leveraging AI expertise to become more efficient drug developers or to offer AI-powered R&D tools as scalable SaaS solutions are likely to thrive. Both industries can learn from each other: biotech can adopt tech's storytelling and direct communication strategies to build public trust and appreciation for scientific innovation, while tech can learn from biotech's rigor, validity, and accuracy, especially in safety-critical applications like medicine. Ultimately, the successful application of AI in biotech promises to democratize scientific discovery, making it faster, cheaper, and more accessible, leading to a future with more groundbreaking medicines.

Action Items

  • Create AI agent runbook: Define 5 sections (purpose, data sources, limitations, validation, use cases) for Benchling AI Deep Research Agent.
  • Audit 10 AI model integration points: Assess for data bias, IP leakage, and regulatory compliance risks within Benchling AI.
  • Implement 3-5 AI simulation tools: Integrate into scientist workflows for predictive modeling of molecule design and experimental outcomes.
  • Measure AI impact on experiment cycle time: Track 5-10 key experiment types to quantify time savings from AI agent usage.
  • Draft AI translation guide: Define 5-7 principles for making AI outputs legible and trustworthy for wet lab scientists.

Related Episodes

Key Quotes

"I worked in a biology lab, I was like really interested in medicine. Coming from the world of software and software developers have amazing tools for working on code and for collaborating, and when I got to the biology lab, I found that scientists had paper notebooks and spreadsheets that would sit on their desktops and like it was terrible. And so it was really hard to work together and I think that was really frustrating for me personally."

Sajith Wickramasekara explains his personal motivation for founding Benchling, highlighting the stark contrast between the sophisticated tools available to software engineers and the rudimentary methods used by scientists in biology labs. This anecdote underscores the problem Benchling aims to solve: modernizing scientific workflows with effective digital tools.

"Making a drug, there's like 9,999 steps in making a drug after you come up with the molecules. You have to find a biologically meaningful target in the body, design a molecule to optimize that molecule, test that molecule in petri dishes and cell lines and animals, various kinds of animals, then eventually you get to where you can take it to a clinical trial and you're testing it in subsequently larger groups of humans. All the while you're figuring out how do I manufacture this thing and develop a process to make it scale economically, safely, with high quality, all while navigating regulatory bodies."

Wickramasekara details the immense complexity and length of the drug development process, emphasizing that Benchling's focus is on organizing and managing the vast, heterogeneous scientific data generated throughout these numerous stages. This illustrates the critical need for a robust system to handle the intricate data landscape of pharmaceutical R&D.

"Biotech has definitely gone through cycles. We're probably like the last couple years are probably like the equivalent of like the dot com bust happening for biotech. It's been a tough time. COVID was sort of a peak when mRNA was this thing that kind of like reopened the world, and you know, there's a lot of generalist money that came in and a lot of exuberance and excitement."

Wickramasekara characterizes the recent period in the biotech industry as a downturn, comparable to the dot-com bust, following a peak driven by the excitement around technologies like mRNA. He notes that this downturn is influenced by various factors beyond the pandemic, including economic shifts and regulatory uncertainties.

"It is probably easier at this point to send things to space or to put people on the moon than it is to get a new medicine approved. And I know $2 billion probably isn't that... when there's that high of a failure rate, it's very difficult for investors to underwrite that. And that was, you know, while we had all these new categories of medicines being invented over the last decade, I think that's like that's important and it's here to stay, but like the industry has to change. The pressure on biotech to be faster and cheaper is just higher than it's ever been before."

Wickramasekara highlights the extreme difficulty and cost associated with drug development, stating it's harder than space exploration and that the high failure rate makes it challenging for investors. He argues that the industry must adapt to increasing pressure for speed and cost-efficiency, especially given the advancements in new medicine categories.

"The AI that wins is going to be the one that people actually use. Give us the temperature check of like what's large pharma and your customer base thinks about AI right now. They've got these AI officers. Oh yeah, there is excitement for sure. There is optimism and belief. I think they're pretty pragmatic though, and I think they're all looking to transform, but they're being methodical."

Wickramasekara emphasizes that practical adoption is key for AI's success in biotech, noting that while large pharmaceutical companies are optimistic and exploring AI, they are proceeding methodically. He suggests that the AI solutions that integrate seamlessly into existing workflows and prove their utility will ultimately prevail.

"I think the biotech and pharma world can learn something from how tech communicates and tells stories. There's the whole go direct wave in tech right now that I really, really resonate with me as a founder. I think like biotech and pharma companies need to tell their stories. Most people, not what I thought you were going to say. Most people, I don't think could name five scientists or five CEOs of pharma companies, but like we could name every single tech CEO."

Wickramasekara suggests that the biotech and pharma industries could benefit from adopting the direct communication and storytelling strategies prevalent in the tech sector. He observes that while tech leaders are widely recognized, prominent figures in biotech and pharma often remain unknown to the general public, hindering broader appreciation for their work.

Resources

External Resources

Books

  • "The Returns to Intelligence" by Dario Amodei - Mentioned as an essay that highlights the high returns to intelligence in scientific progress.

Articles & Papers

  • "The Returns to Intelligence" (Dario Amodei) - Discussed as an essay that highlights the high returns to intelligence in scientific progress.

People

  • Sajith Wickramasekara - Co-founder and CEO of Benchling, discussed for his insights on AI in biotech, drug development, and building interdisciplinary companies.
  • Sarah Guo - Host of the "No Priors: Artificial Intelligence | Technology | Startups" podcast, who is joined by Sajith Wickramasekara for the discussion.
  • Dario Amodei - CEO of Anthropic, mentioned for his essay on returns to intelligence and his company's work in AI.
  • Ashish - Co-founder of Benchling, discussed for his decision to focus entirely on AI development within the company.
  • Milay - A friend and former Benchling employee, mentioned for a compliment regarding Sajith Wickramasekara's understanding of company macro and customer details.

Organizations & Institutions

  • Benchling - A company that provides modern software for scientific progress, serving as a central system of record for biotech R&D and utilizing AI agents.
  • Anthropic - An AI company with a partnership with Benchling, noted for its commitment to science and its AI technology.
  • Johnson & Johnson - Mentioned as a company that partnered with Legend Biotech to develop the successful cancer immunotherapy, Carvykti.
  • Legend Biotech - A Chinese biotech company that partnered with Johnson & Johnson, whose data on Carvykti was promising.
  • Merck - Mentioned as a company that had the courage to invest heavily in the drug Qtruda.
  • Pfizer - Mentioned as a large pharmaceutical company that has partnered with Chinese biotechs.
  • Eli Lilly - Mentioned as a large pharmaceutical company that uses Benchling software.
  • Sanofi - Mentioned as a large pharmaceutical company that uses Benchling software.
  • Moderna - Mentioned as a large pharmaceutical company that uses Benchling software and was involved in mRNA technology during COVID-19.
  • Regeneron - Mentioned as a large pharmaceutical company that uses Benchling software.
  • IceMorf Labs - Mentioned as a cutting-edge AI biotech startup that uses Benchling software.
  • Zera - Mentioned as a cutting-edge AI biotech startup that uses Benchling software.
  • Illumina - Mentioned for providing sequencers that have made sequencing an accessible tool in science.
  • DeepMind - Mentioned as a foundation model company involved in AI for drug discovery.
  • OpenAI - Mentioned as a foundation model company involved in AI for drug discovery.
  • Waymo - Mentioned as an example of an approach to AI autonomy that requires significant investment and time.
  • Tesla - Mentioned as an example of an approach to AI that involves taking steps and making progress.
  • Gilead - Mentioned for its role in developing treatments that have effectively cured HIV.
  • Thermo Fisher - Mentioned as a large company that has succeeded through the systematization of tools in the physical realm.
  • Danaher - Mentioned as a large company that has succeeded through the systematization of tools in the physical realm.

Websites & Online Resources

  • No Priors: Artificial Intelligence | Technology | Startups - The podcast where this discussion took place.
  • show@no-priors.com - Email address for feedback on the podcast.
  • @NoPriorsPod - Twitter handle for the podcast.
  • @Saranormous - Twitter handle for Sarah Guo.
  • @EladGil - Twitter handle for Elad Gil.
  • @sajithw - Twitter handle for Sajith Wickramasekara.
  • @benchling - Twitter handle for Benchling.
  • no-priors.com - Website for the podcast, where transcripts and emails can be found.

Other Resources

  • AI agents - Discussed as tools that can assist scientists, automate work, and recommend experiments.
  • Drug development - Characterized as a costly, time-consuming, and high-failure endeavor that AI aims to improve.
  • Biotechnology - Described as an industry that has gone through cycles, with recent challenges comparable to the dot-com bust.
  • Gene editing - Mentioned as a scientific technology that generated significant investor excitement.
  • Cell and gene therapies - Mentioned as scientific technologies that generated significant investor excitement.
  • RNA - Mentioned as a scientific technology that generated significant investor excitement.
  • mRNA medicines - Mentioned as approved categories of medicines.
  • Platform companies - Discussed in the context of biotech companies being advised to build them in 2021.
  • GLP-1s - Mentioned as a category of drugs that have transformed obesity treatment and are expected to be the best-selling drugs of all time.
  • Contrave - Mentioned as a breakout success in pharma.
  • Qtruda - Mentioned as a drug that exemplifies high returns to intelligence and required courage from Merck.
  • Obesity treatment - Described as a previously unfundable category that has been transformed by GLP-1s.
  • Neurodegenerative diseases - Mentioned as a space that experienced significant failures and investor withdrawal.
  • Alzheimer's - Cited as an example of a disease area with a history of billion-dollar failures.
  • Structure prediction - Mentioned as a tractable problem in AI for biology, with models like AlphaFold.
  • Antibody developability - Mentioned as a tractable problem in AI for biology.
  • Federated learning - Mentioned as a new approach being used in AI for biology.
  • Tune Lab - A project by Recursion Pharmaceuticals that makes internal models available to the scientific ecosystem.
  • AlphaFold - Mentioned as a structure prediction model.
  • Bolts - Mentioned as an open-source model for structure prediction.
  • Manufacturing processes - Discussed as a critical step in drug development that needs optimization for speed and cost.
  • Data transactions - Discussed as a potential future area in biotech, hindered by trust issues with data.
  • Tools - Contrasted with assets as a means of creating value in pharma, with a growing importance in the digital realm.
  • Cloud computing - Mentioned as a norm in tech that was not prevalent in life sciences software when Benchling started.
  • Structured data - Highlighted as a core premise of Benchling's system of record approach.
  • Foundation model labs - Mentioned as sources of AI capabilities for scientists.
  • GPT - Used as a metaphor for advanced AI capabilities without a user-friendly interface for specific domains.
  • AI tutoring - Mentioned as a beneficial application of AI for children's learning.
  • AI coding tools - Mentioned as a personal area of excitement for Sajith Wickramasekara, allowing for rapid creation.
  • ChatGPT - Mentioned as a tool used by Sajith Wickramasekara's mother, representing intuitive technology.
  • Go direct wave - A communication strategy in tech that biotech and pharma companies could learn from.
  • Heroes journey - A narrative style common in tech that could be applied to biotech and pharma.
  • Rigor and validity - Qualities that biotech and pharma have historically excelled at, which tech can learn from.
  • Safety in medicine delivery - Highlighted as a standard that the US has established, which is crucial for patient trust.
  • Capitalism - Mentioned as a topic that can serve as a sorting function when discussing with potential employees.

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.