Ep198: Abbas Kazimi on Computation and Culture for Drug Discovery

What It Covers

Abbas Kazimi, CEO of Nimbus Therapeutics, describes how a 70-person team with no labs used computational chemistry to discover zazositinib, a TYK2 inhibitor sold to Takeda for $4B upfront in 2022. He outlines the capital-efficient hub-and-spoke LLC model, target selection discipline, and culture that enabled two blockbuster exits across 17 years.

Key Questions Answered

• •Hub-and-spoke LLC structure: Nimbus holds assets in separate C-Corps under an LLC parent, allowing pharma partners to acquire single assets without buying the entire platform. This structure enabled a $1.2B Gilead exit for the ACC/NASH program in 2016 and a $4B Takeda deal in 2022, while keeping the core team intact and recycling capital into the next discovery cycle.
• •No-lab CRO model for capital efficiency: Nimbus has operated without internal labs for 17 years, relying on CROs for wet-lab execution while keeping a core team of roughly 70. This dial-up/dial-down workforce model, conceived during the 2008 financial crisis, allowed the company to raise only $63M before generating a $600M payout from the Gilead ACC transaction — approximately a 10x return.
• •Target selection as primary competitive advantage: Before any chemistry begins, Nimbus evaluates targets across five criteria: structural resolvability, well-understood biology with genetic validation, clinical execution feasibility, competitive landscape, and commercial partnering potential. Programs failing any criterion are killed early. This discipline has produced roughly 75% annual portfolio attrition, recycling over $100M back into early discovery by avoiding late-stage failures.
• •Fast-follower strategy with structural differentiation: Nimbus deliberately entered the TYK2 space behind Bristol Myers Squibb, using BMS's clinical progress to refine compound quality. Structural biology work at the atomic level produced 1,500,000-fold selectivity for TYK2 over related JAK family members. Phase 3 data now shows approximately one-third of psoriasis patients achieving complete skin clearance (PASI 100), versus roughly 10-15% for the BMS first-mover compound.
• •Kill programs faster to compound discovery output: Kazimi's strategic shift as CEO focuses on accelerating target onboarding and shortening kill decisions rather than protecting programs. The goal is a steady pipeline of clinical-grade assets rather than one blockbuster every several years. Tighter integration of biology, chemistry, and computation — combined with willingness to take on riskier target classes like molecular glues and degraders — defines the next phase of Nimbus's model.
• •AI as thinking amplifier, not answer engine: Nimbus treats AI and machine learning as tools to extend the reasoning capacity of its scientists, not as autonomous drug designers. The company beta-tests new computational tools from startups and academic partners on live programs rather than building proprietary platforms. This approach keeps the team asset-light on infrastructure while accessing cutting-edge methods, with the human scientific judgment remaining the core differentiator.