Loading CES pipeline...
Available now

CES – Co-culture Efficacy Score

A scoring platform for co-culture screening data

Rapidly quantify and prioritize compound activity in multicellular systems. CES distinguishes effector-mediated modulation from general toxicity, supporting applications across cancer immunotherapy and host-pathogen antiviral screens.

FAQ

General questions

The Co-culture Efficacy Score (CES) is a quantitative framework for measuring drug activity in complex multicellular systems. It integrates cumulative activity and maximal efficacy into a single interpretable score, separating general toxicity from true effector-mediated cytotoxicity.

For a complete breakdown of all pipeline output metrics (including nAUC, Peak, and toxicity flags), refer to the Documentation page

Therapeutic CES: This mode quantifies overall co-culture drug efficacy. It favors compounds that maximize target cell elimination while actively penalizing compounds that compromise effector cell viability, thereby reflecting therapeutic potential in co-culture environments.

Mechanistic CES: This mode isolates immunomodulation by adjusting for compound toxicity against effector cells. It captures drug activity that either enhances or inhibits effector-mediated cytotoxicity by mathematically filtering out artifacts caused by toxicity to the effector cell population.

Applicable to 3-condition setups (co-culture, target monoculture, and effector monoculture).

To upload your co-culture screening data successfully, please format your file according to the CES input requirements. The Documentation page provides detailed guidance on the required columns, file structure, and example input files.
The CES methodology is described in our preprint on bioRxiv. Please see the Cite page for the full citation.

© Copyright 2026 CES. All rights reserved.

Welcome to the CES analysis pipeline!

Upload your data using the sidebar menu. Review the format requirements below before running the analysis.

Raw data format

Upload a .zip archive containing raw plate reader exports, a plate annotation file, and an experimental info file.

  • Plate files: Standard 384-well numeric matrix outputs (.xlsx) inside a folder (use Raw/ folder).
  • Annotation file: Must contain "anno" or "Annotation" in the filename. Maps wells to drugs and concentrations.
  • Info file: Must contain "info" or "breeze" in the filename. Maps screen IDs to filenames. Use Co-culture, Target, and Effector for screen identifiers.
Annotation file format

For raw plate reader signals. Normalized percent inhibition is computed automatically from internal plate controls.

  • Required columns: WELL, PLATE, DRUG, CONCENTRATION, SCREEN_NAME, WELL_SIGNAL.
  • Control anchors: The DRUG column must include BZCL (or POS) and DMSO (or NEG) as negative controls. Co-culture plates may optionally use DMSO2 (or NEG2) for a condition-specific baseline; if absent, DMSO is used.
Processed data format

For data already normalized to percent viability or inhibition.

  • Required columns: DRUG, CONCENTRATION, SCREEN_NAME, RESPONSE.
  • Screen identifiers: Must contain Co-culture and Target (or Mono). An optional Effector (or Control) screen enables toxicity scoring.
Antiviral data format

For evaluating viral inhibition against host cell viability.

  • Required columns: DRUG, CONCENTRATION, SCREEN_NAME, RESPONSE.
  • Screen identifiers: Must contain exactly Co-culture and Mock .
  • Response values: Host Cell Viability (%) for Mock and Viral Inhibition (%) for Co-culture.

© Copyright 2026 CES. All rights reserved.

Technical documentation

1. Overview

Quantifying compound activity in co-culture assays: CES is a computational framework that distinguishes general compound toxicity from true effector-mediated cytotoxicity in multicellular systems. When full 3-condition data (Co-culture, Target, and Effector) is provided, it supports two distinct scoring modes:

  • Therapeutic mode: Penalizes effector-cell toxicity to evaluate overall co-culture efficacy and clinical potential.
  • Mechanistic mode: Adjusts for effector-cell toxicity to isolate purely immunomodulatory (effector-driven) activity.

Two-condition setup (Default): If Effector cell data is omitted, the framework automatically evaluates the baseline delta, computing the direct difference between co-culture and target monoculture cell killing.

Flexible modeling of complex responses: To capture non-linear response landscapes, the framework integrates condition-specific dose-response modeling, co-culture interaction profiling, and quantitative feature extraction. This approach enables the precise identification of both immune enhancers and inhibitors, alongside their optimal effective concentrations.

The pipeline operates in four stages:

  • Stage 1. Quality control and data processing. For Raw / Annotation file uploads, plate-level quality metrics (Z-prime, robust Z-prime, SSMD) are evaluated and viability measurements are normalized to plate-specific controls. Dose-response curves are independently fitted per condition, interpolated onto a uniform dose grid, and combined into co-culture interaction profiles according to the selected scoring mode.
  • Stage 2. Gaussian mixture modeling. Interaction profiles are modeled using a flexible mixture of up to four symmetric Gaussian components to capture non-linear response landscapes. Parameters are estimated by minimizing the sum of squared errors, with differential evolution as a fallback optimizer.
  • Stage 3. Feature extraction and CES computation. The fitted model yields the normalized area under the curve (nAUC), maximal response (Peak), and corresponding effective dose. These features are integrated into the final Co-culture Efficacy Score.
  • Stage 4. Output generation. The pipeline exports the CES, derived pharmacological features, effector toxicity classifications, and interactive dose-response visualizations with model diagnostics.

2. Input data

Upload data as .zip , .csv , .xlsx , or .txt/.tsv . Column names are case-insensitive and common synonyms are automatically mapped (e.g., SCREEN or CONDITION are recognized as SCREEN_NAME ; DRUG_NAME as DRUG ). Required columns and identifiers depend on the chosen pipeline format.

If concentrations are not provided in nanomolar (nM), the DSS values reported in the results table may be inaccurate. All CES-derived metrics remain unaffected.

Accepted column synonyms

The pipeline automatically recognizes the following alternative column headers:

Canonical Also accepted
Annotation / Processed
SCREEN_NAME SCREEN, CONDITION
DRUG DRUG_NAME
CONCENTRATION CONC.(NUM), CONC, DOSE
RESPONSE PERCENT_INHIBITION
Raw (.zip) — Info file
SCREEN_ID SCREEN, SCREEN_NAME, CONDITION
FILE_NAME FILENAME
Raw (.zip) — Annotation file
CONCENTRATION CONC.(NUM), CONC, DOSE
DWell WELL
ProductName DRUG, DRUG_NAME

All column names are case-insensitive.

Drug sensitivity
Raw data (.zip)

A consolidated archive for processing unparsed screens. The uploaded .zip file must contain:

  • Raw plate folder: A folder (use Raw/ folder) containing standard 384-well numeric matrix outputs in .xlsx format.
  • Annotation file: Must contain "anno" or "Annotation" in the filename. Contains well mapping, drug names, and concentration data.
  • Experimental info file: Must contain "info" or "breeze" in the filename. Maps plate filenames to condition identifiers.

Raw plate parsing currently supports 384-well formats. For high-throughput 1536-well plates, please use the Annotation file format.

Condition Mapping: For the screen identifier column in the Info file (e.g., SCREEN , CONDITION , or SCREEN_NAME ), the pipeline expects Co-culture , Target , and optionally Effector . The pipeline automatically recognizes and maps common variations (e.g., mono or stroma to Target; NK, PBMC, or control to Effector).

Example 384-well plate layout highlighting positive (BzCl) and negative controls. DMSO is used for Target and Effector plates. For Co-culture plates, DMSO2 is used as a condition-specific baseline when present; otherwise, the pipeline falls back to DMSO.

Raw/ plate files

Standard 384-well plate reader outputs (.xlsx) stored inside the Raw/ folder of the .zip archive.

Annotation file

Maps wells to drugs and concentrations. The minimum required columns are DWell (or WELL), ProductName (or DRUG), and a concentration column.

Experimental info file

Maps plate filenames to screen condition identifiers. The minimum required columns are FILE_NAME (or FILENAME) and SCREEN_ID (or SCREEN, SCREEN_NAME, CONDITION).

Annotation file

Select this option for raw plate-reader signals formatted in a single file. Normalized percent inhibition is computed using internal plate controls.

  • Required columns: WELL, PLATE, DRUG, CONCENTRATION, SCREEN_NAME, WELL_SIGNAL.
  • Screen identifiers: Use Co-culture , Target , and optionally Effector .
  • Control anchors: The DRUG column must contain positive ( BZCL / POS ) and negative controls ( DMSO / NEG ). Co-culture plates may optionally use DMSO2 or NEG2 for a condition-specific baseline; if absent, DMSO/NEG is used.

Negative control handling: DMSO (or NEG) is the standard negative control for Target and Effector plates. For Co-culture plates, the pipeline uses DMSO2 (or NEG2) if present; otherwise it falls back to DMSO. This means a single DMSO label across all conditions is supported.

Processed data

Select this option if the dataset is already normalized to percent viability or inhibition.

  • Required columns: DRUG, CONCENTRATION, SCREEN_NAME, RESPONSE.
  • Screen identifiers: Use Co-culture and Target baselines. An optional Effector screen enables toxicity scoring.
Antiviral
Processed data

Tailored for evaluating viral inhibition against host-cell viability.

  • Required columns: DRUG, CONCENTRATION, SCREEN_NAME, RESPONSE.
  • Screen identifiers: Must contain exactly Co-culture and Mock (mock-infected host cell viability).
  • Response values: Host Cell Viability (%) for Mock and Viral Inhibition (%) for Co-culture.

3. Outputs

Upon successful processing, the pipeline provides a structured results table and interactive visual diagnostics to facilitate the rapid identification of active compounds.

Plate QC statistics (Raw / Annotation only)
  • Plate_SSMD : Strictly Standardized Mean Difference between positive and negative controls.
  • Signal_Vs_BG : Signal-to-background ratio.
  • Z_Prime & Robust_Z_Prime : Standard assay quality metrics. Values < 0.5 automatically trigger a warning.
Results table metrics
  • CES : Co-culture Efficacy Score. The primary ranking metric that integrates cumulative activity and maximal efficacy into a single value. Positive scores indicate compounds that enhance effector-mediated cytotoxicity (enhancers), negative scores indicate compounds that inhibit it (inhibitors), and values near zero reflect neutral compounds with no measurable modulation.
  • nAUC : Normalized area under the curve of the co-culture interaction profile. Captures the cumulative drug activity across all tested concentrations, providing a measure of overall dose-dependent efficacy rather than activity at a single concentration.
  • Peak : The maximal response amplitude observed in the fitted interaction profile. Represents the strongest modulation achieved by the compound at any concentration, regardless of whether the overall dose-response is monotonic or bell-shaped.
  • Effective_Dose : The concentration at which the Peak response occurs. Compounds achieving high Peak values at lower effective doses are generally more promising candidates, as they exhibit strong modulation at pharmacologically relevant concentrations.
  • IC50 : Half-maximal inhibitory concentration, estimated independently for each experimental condition (Co-culture, Target, and Effector where available). Provides a classical potency reference point to complement the metrics specific to CES.
  • Toxic : Categorical flag identifying compounds with generalized effector-cell toxicity, based on the Drug Sensitivity Score (DSS) of the Effector condition exceeding a user-adjustable threshold. Only available when Effector data is provided. Compounds flagged as toxic may show inflated co-culture efficacy due to direct effector killing rather than true immunomodulation.
  • QC_fit : Quality control status for the underlying dose-response model fits. A flag indicating failure is triggered when the residual error of any condition-specific fit exceeds 30%, indicating that the model may not reliably represent the observed data for that compound.
  • Excluded compounds : Drugs automatically filtered by CES quality control. This occurs when the dose-response profiles across conditions exhibit biologically implausible behavior, such as concordant proliferative signals (strongly negative inhibition in both monoculture and co-culture), indicating assay artifacts rather than genuine pharmacological activity. Excluded compounds are retained in the results table for transparency but are not assigned a CES value.

Extended download: The exported file (.csv/.xlsx) expands the visual table to include individual condition Drug Sensitivity Scores (DSS), model error metrics, and dynamically scaled raw dose-response profiling measurements.

Interactive visualizations
  • Plate QC (Raw / Annotation only): Cross-screen boxplots displaying the distribution of raw well signals for each internal control type, alongside spatial scatter plots of signal intensity across plate columns. Together these reveal systematic issues such as edge effects, plate drift, or poor control separation that could compromise downstream scoring.
  • CES distribution: Ranked bar chart ordering all scored compounds from highest to lowest CES. Enhancers (positive CES) are colored red, inhibitors (negative CES) blue, and neutral compounds grey, providing an immediate visual summary of the overall screening landscape.
  • CES vs. Effective Dose: Scatter plot mapping each compound by its effective dose (x-axis, log-scaled) against its CES (y-axis). Compounds in the upper-left quadrant combine strong efficacy with low effective concentrations, highlighting the most promising candidates for further investigation.
  • Dose-response modeling: Per-compound interactive curve fits displaying the modeled dose-response across all experimental conditions. Raw data points are overlaid on the fitted curves, allowing visual inspection of model quality, condition-specific response differences, and the pharmacological behavior driving the integrated co-culture interaction profile.

4. Contact

For questions regarding the analytical framework or web application, please contact:

© Copyright 2026 CES. All rights reserved.

Citing CES

If you use CES in your research, please cite our preprint:

Dias, D. et al. (2026). Selective scoring of drug effects in multicellular co-culture systems.

bioRxiv , doi: 10.64898/2026.05.20.726737

View on bioRxiv

Source code and software

The CES computational framework and this interactive platform are entirely open-source. Explore the methodology or deploy it locally.

View repository on GitHub

© Copyright 2026 CES. All rights reserved.