High-content Profiling in Oesophageal Adenocarcinoma for Identification and Mechanistic Characterisation of Small Molecules to Inform Preclinical Development and Patient Stratification


Session type:

Rebecca Hughes1,Richard Elliot1,Ashraff Makda1,Robert O'Neill1,Ted Hupp1,Neil Carragher1
1University of Edinburgh



Oesophageal adenocarcinoma (OAC) is a highly heterogeneous disease, dominated by large-scale genomic rearrangements and copy number alterations. Such characteristics have hampered therapeutic target discovery and patient stratification. We describe an alternative phenotypic-led drug discovery platform using an image-based high-content profiling assay to classify drug mechanism-of-action (MOA). We have applied this across a panel of OAC and tissue matched non-transformed cell lines as a novel strategy for drug discovery.


We applied the CellPainting assay using multiplexed fluorescent dyes to profile the phenotypic response of our panel of OAC lines to reference compounds in distinct mechanistic classes. Approximately 1000 features were extracted from the images and used to train a machine-learning model capable of predicting MOA for uncharacterised compounds. Combining this assay and the prediction model, we screened 16,500 small molecules from diverse chemical libraries (BioAscent and CRUK Therapeutics Discovery Labs) to allow both the identification of hits and assignment of MOA, thus informing further preclinical development and future patient stratification hypotheses.


Of 13 STR verified lines assessed, six OAC cell lines, CP-A (a Barrett’s oesophageal cell line), and EPC-2 (a squamous oesophageal cell line) were suitable for image-based profiling and also represent the genetic diversity of OAC.

Our unbiased phenotypic approach allowed accurate classification of MOA by machine-learning, at 80-90 % accuracy across the panel of 8 lines.  

We have further prioritised hit compounds which selectively target the growth or survival of specific OAC cell lines over non-transformed oesophageal cells, several of which represent novel MOA distinct from our reference compound set.


Our high-content OAC assay has proven effective in the identification and mechanistic characterisation of several small molecules. Further validation of hit compounds across our suite of disease-relevant 2D and 3D OAC cell models provides an evidence-led platform to stimulate new drug discovery programs in OAC.