In cell analyzer systems

For IF assays, ChiPS4 cells were seeded on μ-Slide 8-well (Ibidi) or 96-well plates suitable for High Content Microscopy (Nunc). Standard IF procedure was used where appropriate. Briefly, following treatments, cells were washed with PBS, fixed with 4% formaldehyde, blocked in 5% BSA in PBS-T, and incubated with primary followed by Alexa-conjugated secondary antibodies. Cell Painting was performed as described by Bray et al. (29 (link)). Imaging and subsequent analysis were performed using IN Cell Analyzer systems (GE Healthcare) and Spotfire (Tibco). Main measures extracted from the Cell Painting assay data set are: area of nuclei (μm²) calculated as a number of pixels in nucleus, multiplied by the area per pixel; nuclei form factor calculated as 4∗π∗Area/Perimeter² is a measure of circularity (values between 0 and 1 for a perfectly circular object); nuclei: FITC texture correlation is a measure of relative roughness of the image within the nucleus correlating to the nucleolar staining.

We employed a subset of SRS images in the fixed lung cancer cells (A549, from ATCC) dataset as one of our pre-trained data sources. These data sets were acquired simultaneously using image software by collecting the SRS signals from lock-in amplifiers and fluorescence signals from photomultiplier tubes [36 (link)]. For the fluorescence signals, all dyeing schemes are based on the standards provided that three different color fluorescent dyes were used to label and track the nucleus, mitochondria, and endoplasmic reticulum, respectively. The optical cell images with 512 × 512 pixels were obtained at a dwell time of 4 μs.
Another trained source of data we employed is the dataset cell images, which are acquired using GE’s IN Cell Analyzer systems [53 (link)]. These datasets were applied to test different deep learning methods in the work and evaluate their performance.