Large-scale perturbational profiling is time and cost-intensive, so it’s crucial to compare the capabilities of different technologies to help plan and design experiments efficiently.
High-throughput screening approaches like the Cell Painting morphological profiling assay and L1000 gene expression profiling are powerful tools that generate detailed phenotypic and molecular insights into cellular responses for companies and researchers performing drug discovery, toxicology, or genetic perturbation screens.
However, while these fundamentally different technologies are useful screening tools in their own right, can combining these approaches lead to more refined or complementary predictions of mechanisms of action, off-target effects, or toxicity following drug treatment?
Read on to find out.
Cell Painting versus L1000: What’s the difference?
Cell Painting is a high-content image-based assay that uses six fluorescent dyes to label distinct subcellular compartments like the nucleus and endoplasmic reticulum (Bray et al., 2016). Machine learning algorithms then analyze these images to extract several thousand morphology measurements from each cell, some of which might change in response to perturbations, such as drug treatments or CRISPR screens.
The output from this image analysis can robustly predict mechanisms of action, toxicity, or off-target effects of drugs. However, Cell Painting gives no insight into changes at the gene expression level, meaning crucial expression pathways, important genic targets, or subtle mechanisms of action might be missed.
In contrast, L1000 uses Luminex beads, each tagged with a different “landmark” gene sequence that hybridizes to and assesses the mRNA abundance of 978 genes selected to represent the diversity of pathways and processes in human cells (Subramanian et al., 2017). The expression of a further 11,350 genes can be inferred computationally despite not being directly measured.
Unlike Cell Painting, L1000 doesn’t provide any information about cellular morphology, as it’s limited to detecting gene expression alterations that don’t necessarily translate to functional changes in cellular state.
Does combining Cell Painting and L1000 provide useful complementary information?
In 2022, researchers from the Broad Institute of MIT and Harvard sought to address these key limitations by assessing how data and phenotypic insights derived from morphological profiling using Cell Painting compared to gene expression profiling with L1000 (Way et al., 2022).
Their goal was simple: to work out the extent of overlapping versus distinct information captured by each assay for the mechanistic understanding of drugs.
The researchers took a comprehensive screening approach similar to toxicology or drug discovery screens performed in industry to generate dose-response data. They treated human A549 lung cancer cells with over 1,300 small molecules from the Drug Repurposing Hub across six doses with between three and five replicates, then performed the Cell Painting and L1000 assays on each.
Here’s what they found:
- Cell Painting was more reproducible between replicates (57%-83% replicability across doses) compared to L1000 (16%-35%). However, Cell Painting was more sensitive to batch effects than L1000, but this could be corrected for meaningful comparisons.
- Although L1000 had lower levels of reproducibility, it captured more independent features, suggesting a broader range of molecular information. This increased diversity indicated more potential scope for interesting biological findings in response to treatments.
- While L1000 captured more overall features, Cell Painting morphology profiles still captured more diverse cell states.
- Cell Painting and L1000 detect different drug mechanisms, reinforcing the complementarity of combining their outputs for drug discovery or toxicology pipelines.
Overall, the researchers found that combining the two methods could detect 69% of assayed mechanisms of action, providing broad levels of coverage. 19% of mechanisms of action were detected by Cell Painting only, 24% by L1000 only, and 27% by both assays.
L1000 excelled at detecting MAPK and heat shock protein inhibitors at the gene target level, whereas Cell Painting was more precise for Aurora kinase inhibitors, Polo-like kinase inhibitors, and BRD4.
For drug discovery studies aiming to generate as much phenotypic and molecular insight as possible, it’s clear that combining different high-throughput screening technologies can be a powerful option.
What’s next for Cell Painting and L1000?
The Broad Institute researchers have now taken Cell Painting to the next level. The publicly available Joint Undertaking for Morphological Profiling (JUMP) Cell Painting Gallery contains images of cells responding to more than 140,000 perturbations like drug treatment or gene modifications, alongside other Cell Painting data sets (Weisbart et al., 2024).
Cell Painting-based bioactivity prediction was also recently shown to boost high-throughput screening hit rates and compound diversity (Haslum et al., 2024). The authors state that “the approach has the potential to reduce the size of screening campaigns, saving time and resources, and enabling primary screening with more complex assays.”
In 2025, another group of researchers reported the first-ever genome-scale perturbation atlas of morphology phenotypes in human cells (Ramezani et al., 2025). They combined a Cell Painting-based assay with massively parallel optical pooled CRISPR screens followed by in situ sequencing by synthesis to assign perturbations to cells. The assay, called PERISCOPE, profiled the effects of over 20,000 single gene knockouts in over 30 million cells. While the dataset is undoubtedly robust, the assay provides no transcriptional information, leaving any gene expression changes in the dark.
Similarly, the L1000 assay has shown utility in drug discovery, thanks partly to large-scale publicly available databases like the Connectivity Map. For instance, researchers recently developed a novel algorithm called PRnet trained on an L1000 dataset consisting of 883,269 transcriptional profiles from 82 cell lines perturbed by 175,549 biologically active compounds (Qi et al., 2024).
The model predicts transcriptional responses to novel drug treatments and allows in-silico drug screening for diseases based on gene signatures. It found and validated novel drug candidates against small-cell lung and colorectal cancer.
MERCURIUS™ DRUG-seq: a better choice than L1000?
Combining morphological profiling assays like Cell Painting with gene expression profiling assays like L1000 can clearly help advance drug discovery pipelines. However, L1000 has multiple core limitations, as Way et al. show.
For instance, L1000 directly measures only 978 genes and computationally infers the expression of another 11,350, which is problematic if a broader overview is required. Way et al. also found that L1000 had low reproducibility between replicates and missed many mechanisms of action detected by Cell Painting.
Each of these issues is potentially due to the limited direct expression measurements and inaccuracies or uncertainty generated by L1000’s computational inference approach. With new technologies, like MERCURIUS™ DRUG-seq, researchers can harness next-generation sequencing to generate unbiased transcriptome-wide data directly for over 15,000 genes at relatively shallow sequencing depths of 1.5 million reads per sample at low cost. The direct gene expression measurements made with MERCURIUS™ DRUG-seq could limit reproducibility issues as the assay has excellent concordance between replicates and it’s possible that transcriptome-wide approaches like MERCURIUS™ DRUG-seq would provide a more thorough assessment and identification of mechanisms of action.
The ultra-high-throughput MERCURIUS™ DRUG-seq assay allows users to treat cells with drug compounds and lyse the cells directly in 384-well plates. Thanks to the early barcoding step, samples can then be combined in a single tube and proceed with library preparation and sequencing without any prior RNA extraction needed. This ultra-scalable technology leads to unbiased transcriptomics readouts for low-cost drug screening, genetic perturbation studies, or toxicology studies.
Find out more about how MERCURIUS™ DRUG-seq can help boost your drug discovery pipeline by scheduling a call with one of our experts here.
References
- Bray, M.A., et al. Cell Painting, a high-content image-based assay for morphological profiling using multiplexed fluorescent dyes. Nature Protocols, 11(9), pp.1757-1774.
- Fredin Haslum, J., et al. Cell Painting-based bioactivity prediction boosts high-throughput screening hit-rates and compound diversity. Nature Communications, 15(1), p.3470.
- Qi, X., et al. Predicting transcriptional responses to novel chemical perturbations using deep generative model for drug discovery. Nature Communications, 15(1), pp.1-19.
- Ramezani, M., et al. A genome-wide atlas of human cell morphology. Nature Methods, pp.1-13.
- Subramanian, A., et al. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles. Cell, 171(6), pp.1437-1452.
- Way, G.P., et al. Morphology and gene expression profiling provide complementary information for mapping cell state. Cell systems, 13(11), pp.911-923.
- Weisbart, E., et al. Cell Painting Gallery: an open resource for image-based profiling. Nature Methods, 21(10), pp.1775-1777.
