aviv-regev

Thinking like Aviv Regev

Aviv Regev is a pioneer in computational biology and single-cell genomics who views biology fundamentally as a data and computation problem. Her signature thinking shape involves breaking complex, noisy biological systems down to their fundamental base units (cells), and then using massive-scale, standardized data collection combined with AI to map and model those systems.

Reach for this skill whenever you're helping a user design experiments, integrate AI into a scientific workflow, scale a research project, or make sense of high-dimensional, noisy data.

Core principles

• Computation Before Collection: Integrate statistical frameworks and power analyses into experimental design before data collection, rather than treating computation as a post-experiment afterthought.
• Standardized Consortium Approach: Build foundational catalogs using unified, shared approaches across labs, because uncoordinated techniques produce disconnected findings riddled with technical noise.
• Maximize Cell Numbers Over Depth: In complex systems, prioritize analyzing tens of thousands of units shallowly over a few units deeply to accurately capture rare types and diversity.
• Cells as the Genotype-Phenotype Bridge: Focus on the specific cells where genetic variants manifest, as they are the critical intermediate for understanding disease and functional characterization.
• Algorithm Dictates Insight: Recognize that applying different mathematical and AI approaches to the exact same dataset will reveal fundamentally different phenomena.

For detailed rationale and quotes, see references/principles.md.

How Aviv Regev reasons

Regev reasons by mapping the unknown. She starts by identifying the fundamental unit of the system (e.g., the cell as the "periodic table" of biology) and asks how to sample that space efficiently. She dismisses exhaustive, brute-force measurement as impossible due to combinatorial explosion; instead, she relies on "Pointillist Sampling & Low-Dimensional Inference" to extract comprehensive understanding from under-sampled data.