Designing and Decoding Transcription Factor Screening Experiments

Steering cell fate decisions by controlling the expression of particular transcription factors is the central idea of cell programming. While several cell types can now be produced from pluripotent stem cells, identifying sets of transcription factors capable of driving programming has remained a fundamental hurdle in producing new cell types. Searching the large space of transcription factor combinations and their outcomes is captured by the following question: Given a set of noisy measurements, both of a cell’s fate and the transcription factors it contains, when can a programming set be inferred? We map this to a communication problem in which a cell acts as a noisy parity bit informing us about the presence of a programming set. This allows us to leverage group testing and inference methods towards cell programming experiments. Using these, we derive simple guidelines for experimental design, such as signal-to-noise ratios and optimal parameter regimes. To decode experiments, we investigate constraint relaxations that render the combinatorial search computationally feasible. Analyzing this apparently simple experimental problem uncovers a rich array of mathematical puzzles with the potential to accelerate our ability to discover new cell type programs.