Reproducible image-based profiling with Pycytominer

Abstract

Advances in high-throughput microscopy have enabled the rapid acquisition of large numbers of high-content microscopy images. Whether by deep learning or classical algorithms, image analysis pipelines then produce single-cell features. To process these single-cells for downstream applications, we present Pycytominer, a user-friendly, open-source python package that implements the bioinformatics steps, known as "image-based profiling". We demonstrate Pycytominer's usefulness in a machine learning project to predict nuisance compounds that cause undesirable cell injuries.

Figure 1.. The standard image-based profiling experiment and the role of Pycytominer.

(A) In the experimental phase, a scientist plates cells, often perturbing them with chemical or genetic agents and performs microscopy imaging. In image analysis, using CellProfiler for example, a scientist applies several data processing steps to generate image-based profiles. In addition, scientists can apply a more flexible approach by using deep learning models, such as DeepProfiler, to generate image-based profiles. (B) Pycytominer performs image-based profiling to process morphology features and make them ready for downstream analyses.

Figure 2.. Model performance and evaluation with JUMP data.

(A) Our pycytominer-based workflow to process publicly-available data and to train a machine learning model to predict cellular injury. (B) Precision and recall scores for predicting various cellular injuries, comparing the not shuffled model (solid lines) with the shuffled model (dashed lines) across distinct injury types and data splits, with blue indicating the training set and red indicating the test set. The F1 scores for each injury represent only the testing and training datasets with the not shuffled model. (C) Confusion matrices assessing the model’s predictive performance across training, testing, and holdout data. (D) Cytoskeletal injury probability distribution generated by the shuffled (red) and non-shuffled (blue) models. The three groups represent the ground truth wells (top), wells predicted to have cytoskeletal injury (middle), and wells not predicted to have cytoskeletal injury (bottom).