Compound Functional Prediction Using Multiple Unrelated Morphological Profiling Assays

Rose, France; Basu, Sreetama; Rexhepaj, Elton; Chauchereau, Anne; Nery, Elaine Del; Genovesio, Auguste

doi:10.1177/2472630317740831

Cited by 18 publications

(23 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One reason for this observation may be the fact that all reporter cell lines expressed BFP nuclear and cytoplasmic segmentation markers, and in many cases these on their own were sufficient to distinguish MoAs with reasonable accuracy. The predictive power of nuclear and cytoplasmic markers has recently been noted in both live-and fixed-cell phenotypic screens using small sets of annotated reference compounds [4,31] and our results extend this finding.…”

Section: Discussionsupporting

confidence: 86%

Tales of 1,008 Small Molecules: Phenomic Profiling through Live-cell Imaging in a Panel of Reporter Cell Lines

Cox

Jaensch

Waeter

et al. 2020

Preprint

View full text Add to dashboard Cite

Phenomic profiles are high-dimensional sets of readouts that can comprehensively capture the biological impact of chemical and genetic perturbations in cellular assay systems. Phenomic profiling of compound libraries can be used for compound target identification or mechanism of action (MoA) prediction and other applications in drug discovery. To devise an economical set of phenomic profiling assays, we assembled a library of 1,008 approved drugs and well-characterized tool compounds manually annotated to 218 unique MoAs, and we profiled each compound at four concentrations in live-cell, high-content imaging screens against a panel of 15 reporter cell lines, which expressed a diverse set of fluorescent organelle and pathway markers in three distinct cell lineages. For 41 of 83 testable MoAs, phenomic profiles accurately ranked the reference compounds (AUC-ROC ≥0.9). MoAs could be better resolved by screening compounds at multiple concentrations than by including replicates at a single concentration. Screening additional cell lineages and fluorescent markers increased the number of distinguishable MoAs but this effect quickly plateaued. There remains a substantial number of MoAs that were hard to distinguish from others under the current study's conditions. We discuss ways to close this gap, which will inform the design of future phenomic profiling efforts.

show abstract

Section: Discussionsupporting

confidence: 86%

Tales of 1,008 Small Molecules: Phenomic Profiling through Live-cell Imaging in a Panel of Reporter Cell Lines

Cox

Jaensch

Waeter

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…One reason for this observation may be the fact that all reporter cell lines expressed BFP nuclear and cytoplasmic segmentation markers, and in many cases these on their own were sufficient to distinguish MoAs with reasonable accuracy. The predictive power of nuclear and cytoplasmic markers has recently been noted in both live- and fixed-cell phenotypic screens using small sets of annotated reference compounds 4 , 32 and our results extend this finding.…”

Section: Discussionsupporting

confidence: 86%

Tales of 1,008 small molecules: phenomic profiling through live-cell imaging in a panel of reporter cell lines

Cox

Jaensch

Waeter

et al. 2020

Sci Rep

View full text Add to dashboard Cite

or will not assign the molecule at all if reference compounds with similar phenomic profiles are not present in the experiment. For poly-pharmacological compounds, morphological and transcriptional readouts reflect the actions on multiple, functionally unrelated targets that are convolved into a complex phenotype, and their activity on specific and relevant targets must therefore be deconvolved from these phenomic profiles, which is far from trivial. Deconvolution can be achieved using supervised machine learning models to predict target activity 7. This approach requires that many of the phenomically profiled compounds have also been tested in the targetor MoA-specific assays it aims to cover. These compounds are used as training examples to identify the most informative phenomic features to predict activities in each assay of interest. Consequently, the data scale required for supervised deconvolution can be prohibitive. A continuing challenge in drug discovery is the design of an affordable set of phenomic profiling assays with maximal coverage of therapeutic targets. Ideally, this assay set should be able to document a phenotypic response towards at least the vast majority of the 549 human proteins targeted by 999 FDA-approved small-molecule drugs for human disease 12. Until recently, most studies that use phenomic profiling have been restricted in scope, typically using fewer than 100 reference compounds annotated to a limited number of targets 1,2,4,13-15 and are not sufficiently broadly representative to inform design principles for generic sets of phenomic profiling assays. In the current study, we used a nearest-reference approach to explore how various screening parameters affect the ability to distinguish MoAs from each other. We used gene-editing to create a panel of 15 reporter cell lines by introducing different combinations of 12 blue fluorescent protein (BFP), green fluorescent protein (GFP), or red fluorescent protein (RFP)/FusionRed signalling pathway and organelle markers into the A549, HepG2, and WPMY1 cell backgrounds, and profiled these reporter cell lines in live-cell, high-content imaging screens against a library of 1,008 small molecules, manually annotated with 218 unique MoA descriptors, at four concentrations. Results Library of 1,008 reference compounds and 169 natural products. We assembled a set of 1,008 well-characterized reference compounds, composed of FDA-approved drugs and commercially available tool compounds, and manually annotated each compound with one or more MoA descriptors using publicly available information from vendor compound catalogues, chemical databases, and large-scale target annotation projects 12,16-18 (Supplementary Table S1). In total, 218 unique MoA descriptors were assigned to the reference compound set. Of the 1,008 reference compounds, 829 (82%) were labelled with only a single MoA descriptor. Of the 218 MoAs, 132 (61%) were assigned to ≥ 3 co-annotated compounds and 92 (42%) were assigned to ≥ 5 co-annotated compounds (Supplementary Fig. S1). To increase...

show abstract

“…In a pooling analysis such as [38], the cells of the respective cell lines are first normalised and then grouped across drugs to increase the amount of available data. An ensemble approach such as in [29] creates models for each cell line and aggregates their individual predictions. This approach has the additional advantage of allowing different imaging modalities of fluorescent markers.…”

Section: Multi-cell-line Analysismentioning

confidence: 99%

Domain-invariant features for mechanism of action prediction in a multi-cell-line drug screen

Boyd

Pinheiro

Nery

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

High Content Screening is an important tool in drug discovery and characterisation. Often, high content drug screens are performed on one single cell line. Yet, a single cell line cannot be thought of as a perfect disease model. Many diseases feature an important molecular heterogeneity. Consequently, a drug may be effective against one molecular subtype of a disease, but less so against another. To characterise drugs with respect to their effect not only on one cell line but on a panel of cell lines is therefore a promising strategy to streamline the drug discovery process. The contribution of this paper is twofold. First, we investigate whether we can predict drug mechanism of action (MOA) at the molecular level without optimisation of the MOA classes to the screen specificities. To this end, we benchmark a set of algorithms within a conventional pipeline, and evaluate their MOA prediction performance according to a statistically rigorous framework. Second, we extend this conventional pipeline to the simultaneous analysis of multiple cell lines, each manifesting potentially different morphological baselines. For this, we propose multitask autoencoders, including a domain-adaptive model used to construct domain-invariant feature representations across cell lines. We apply these methods to a pilot screen of two triple negative breast cancer cell lines as models for two different molecular subtypes of the disease.

show abstract

Compound Functional Prediction Using Multiple Unrelated Morphological Profiling Assays

Cited by 18 publications

References 24 publications

Tales of 1,008 Small Molecules: Phenomic Profiling through Live-cell Imaging in a Panel of Reporter Cell Lines

Tales of 1,008 Small Molecules: Phenomic Profiling through Live-cell Imaging in a Panel of Reporter Cell Lines

Tales of 1,008 small molecules: phenomic profiling through live-cell imaging in a panel of reporter cell lines

Domain-invariant features for mechanism of action prediction in a multi-cell-line drug screen

Contact Info

Product

Resources

About