Is SaaS a good business model for drug‑discovery companies?
a year ago
- #AI
- #Drug-Discovery
- #SaaS
- The 2024 Nobel Prize in Chemistry recognized advances in computational protein design, sparking interest in foundation models for biology.
- Proteins play a crucial role in drug effectiveness, and AlphaFold2 demonstrated deep learning's potential in predicting protein structures.
- Companies like EvolutionaryScale and Profluent are developing powerful generative models for protein science.
- Licensing AI models as SaaS to pharma is considered a weak business model due to misalignment with pharma's core needs.
- Pharma's main challenge is the 'Valley of Death' between molecule discovery and clinical success, where current models fall short.
- Drug developers focus on ADME, PK/PD, and CMC, areas where foundation models currently lack reliability.
- The failure of Merck’s verubecestat highlights the gap between in-silico predictions and clinical outcomes.
- AI-first startups like Recursion and Insitro are pivoting to owning their molecules to control more of the value chain.
- Open-source models, like the replication of AlphaFold3, threaten the pricing power of SaaS licenses.
- Pharma prioritizes clinical success over molecule discovery costs, making SaaS models less attractive without downstream risk reduction.