Quantifying the interplay between single cell morphology and proteomic signature

Proteins quantification in a spatially resolved tissue is a growing technology, allowing to bridge the gap between tissue organization studies and molecular profiling assays. Spatially resolved omics make it possible to characterize the single cell`s molecular profile while maintaining the tissue spatial organization. This in turn allows to attribute spatial characteristic of a tissue when modeling single cell molecular variation. This is a huge leap towards understanding the relations between a tissue structure and function. Our hypothesis is that using morphological features that are extracted from these images could significantly improve the prediction of single cell proteomic signature. Here, we use publicly available spatially resolved triple negative breast cancer (TNBC) tissues from 41 patients. In each tissue 36 proteins were quantified using multiplexed ion beam imaging by time-of-flight (MIBI-TOF). The protein expression was quantified for each cell, and this quantity is the cell`s proteomic signature. We demonstrate that by incorporating morphological features to our model we can see improvement of the prediction as compared to a baseline model. Our preliminary results indicate morphological-dependent protein expression levels in different cell types. This shows that morphological features of a cell given the cell-type are valuable for predicting cell state in TNCB tissue. Furthermore, we found that some proteins are more associated to cell`s morphology than others, namely proteins that are used as markers for cell status (dsDNA, H3K27me3, H3K9ac). We propose morphological features to be considered when modeling single cell state in a spatially resolved dataset, to improve cell state predictions.