# Phenotypic profiling of small molecules using cell painting assay in HCT116 colorectal cancer cells

## TL;DR

This study uses cell painting to group small molecules by their effects on colorectal cancer cell morphology, revealing shared phenotypes beyond traditional classifications.

## Contribution

The study demonstrates that cell painting can identify convergent phenotypic signatures of small molecules that go beyond their known mechanisms of action.

## Key findings

- 18 distinct phenotypic clusters were identified from 196 small molecules using cell painting data.
- Compounds with different mechanisms of action often converged on similar cellular phenotypes.
- Phenotype-driven screening revealed unexpected relationships among chemically diverse compounds.

## Abstract

Understanding how small molecules affect cellular morphology is essential for exploring their mechanisms of action (MoA) and identifying new therapeutic candidates. In this study, the Cell Painting Assay (CPA) was used to profile 196 small molecules in HCT116 colorectal cancer cells. By applying t-distributed stochastic neighbor embedding (t-SNE) followed by density-based clustering, 18 distinct phenotypic clusters were identified based on similarities in the quantitative morphological profiles generated from Cell Painting data. Although it was initially hypothesized that clustering would reflect known MoAs, most clusters showed only partial overlap with target-based classifications. Instead, compounds from different MoA classes converged on similar cellular phenotypes, suggesting common downstream effects or shared stress responses. Notably, compounds affecting DNA replication, mitosis, or transcriptional control appeared across multiple clusters, indicating functional diversity within morphologically similar groups. Clusters enriched with mTOR/PI3K inhibitors, spindle poisons, or transcriptional CDK blockers exhibited well-defined phenotypes, supporting the robustness of the assay. In contrast, compounds inducing more subtle phenotypes formed distinct micro-clusters, highlighting the method’s sensitivity. Overall, this study demonstrates that CPA can capture convergent phenotypic signatures that extend beyond target-based classification. These findings underscore the value of phenotype-driven screening for the functional annotation of chemical compounds and may help uncover unexpected relationships among molecules with diverse biological activities.