Computer Science, ETH Zurich
Meeting: Neural Optimal Transport for Inferring Single-Cell Responses to Perturbations
Predicting heterogeneous cell responses to chemical or genetic perturbations at the level of single cells is crucial for deciphering molecular processes and obtaining a better understanding of function and disease. While the advent of single-cell high-throughput methods may make this task look easy, their destructive nature prevents us from observing the same cell before and after a perturbation. To predict a patient's response to different treatments, one needs to thus re-align unpaired snapshots of cell populations pre- and post-treatment and predict for each cell its corresponding perturbed state after treatment. In addition, while massively parallel high-resolution methods such as Perturb-Seq allow phenotyping in an unprecedented resolution, their scale and randomized nature pose additional challenges and requirements to machine learning algorithms for modeling perturbation responses. In this talk, I demonstrate how we can use neural optimal transport (OT) methods to solve these puzzles and predict treatment responses optimally on the single-cell level. These novel deep learning approaches inspired by OT theory not only achieve a new state-of-the-art with substantial quantitative improvements to prior works but also open new frontiers in a current large-scale clinical study to predict treatment responses of unseen patients. Beyond that, neural optimal transport schemes can be extended to various experimental and biological settings, and for example, adapt to the randomized and composite nature of large-scale profiling technologies, predict responses to combination therapies, or model high levels of apoptosis and proliferation emerging in cellular perturbation responses.
Department of Cellular and Tissue Genomics - Oncology, Genentech
Primer: Analytical challenges and opportunities for studying cell state transitions at the single cell level
Cell state transitions are at the core of biology: they determine how cells differentiate into each cell type in our body, respond to their environment, change in disease and are altered by treatments. Understanding the rules that govern such cell state decisions is at the heart of learning how living organisms work, and is key for designing long-lasting treatments to reverse disease one cell at a time. Recent advances in single-cell genomics have unlocked the potential to dissect cell state transitions with unprecedented scale and resolution. First, single-cell genomics allows deep characterization of cell states as high-dimensional phenotypes, without needing to decide a priori what aspects of the cell state to experimentally quantify. Second, pooled screens coupled with single-cell readouts enable dissecting the roles of increasingly large numbers of perturbations in parallel and in a directly causal direction, allowing us to learn the latent manifold of cell states, as well as the connections between such states as measured by experimental perturbations. These fundamental advances in our quantitative and predictive understanding of cell state transitions crucially depend on the development of computational models that are able to extract generalizable principles for how cells work. In this primer, I will describe the analytical challenges and opportunities for inferring cell state transitions from a variety of different single-cell screening modalities, including genetic -, chemical - and treatment-based perturbations. I will focus on key questions in this domain, including accurately quantifying the effects of individual perturbations, predicting the additivity and non-linearity of combinatorial perturbations, predicting how a perturbation’s effect depends on a cell’s existing cell state and modeling how collections of cells respond to perturbations. Finally, I will summarize how advances in computational approaches may synergize with our ever-expanding experimental ability to measure increasingly high-resolution cell states across multiple modalities, as well as time and space.