High-dimensional morphological profiling captures cellular changes in response to external stimuli such as cytokines, chemokines, growth factors, antibodies and a wide range of other perturbations. Recursion built a comprehensive library of 434 immune stimulants and embarked on a large-scale effort to profile cellular response to not only demonstrate that cellular images are sufficient to appropriately cluster immune perturbations by function but to rapidly employ these states for high-throughput drug screening applications.
Cells constantly sense their microenvironment to maintain homeostasis and respond to infection and malignancy. In response, cells communicate locally and systemically with a highly diverse array of secreted factors. Both underfunctioning and overfunctioning abnormalities in this process drive a wide range of diseases, and intervention in this process has yielded some of the most successful drugs available. While these processes are typically studied in isolation, a sufficiently high-throughput and high-dimensional approach unlocks the potential for accurate modeling of the immune system and rapid deployment of phenotypic screening programs. Recursion highlights its work in immunity in a recent preprint where four primary human cell types were treated with the same comprehensive immune stimulant library. The images and high-dimensional embeddings from one of these cell types (HUVEC) are available here.
The library components are annotated by general function in the preprint, which serves as a guide for expected high-dimensional similarity (for example, type-I IFNs generate similar morphology).
RxRx2 demonstrates both the great variety of morphological effects soluble factors have on HUVEC cells and the consistency of these effects within groups of similar function. Through RxRx2, researchers in the scientific community will have access to both the images and the corresponding deep learning embeddings to analyze or apply to their own experimentation. The embeddings are 128-dimensional vectors with one vector for each image and come from Recursion’s internal model trained on additional cell types and perturbation modalities not otherwise released. We provide these embeddings to more easily enable researchers without significant compute resources to still explore and uncover insights from this data. Scientific researchers can use the data to further demonstrate how high-content imaging can be used for screening immune responses and identification of functionally-similar factor groups.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
The RxRx2 dataset is closely related to datasets previously released by Recursion, although there are some key differences. For ease of comparison and understanding, we provide the following table highlighting the primary differences:
A CSV containing the experiment design, e.g. what cell type and treatment are in each well. The schema is provided in the README.
A large CSV file containing all of the deep learning embeddings for each image.