DENV-assay-TS-DR Dataset

This dataset concatenates time-series and dose-response data for dengue virus (DENV) antiviral assays. It includes the high-content images from each of the assays, with detailed measurements of antiviral activity at specific timepoints used to elucidate mode of action of potential inhibitors. The notebook shows a workflow for integrating time-series (TS) and dose-response (DR) data from high-content imaging (HCI) antiviral assays and scoring dose-response data from hits using a machine learning model trained on the time-series data.

Overview

The goal of this pipeline is to extract meaningful insights by analyzing both time-series data (captured at multiple timepoints post-infection) and dose-response data (compound concentrations).

The notebook provides an end-to-end analysis, including:

• Feature extraction (using CellProfiler)

• Batch correction (using Harmony)

• Model training and evaluation (with H2O AutoML and standalone XGBoost classifier)

• Scoring of dose-response plates

• Visualization with UMAP, t-SNE, and PHATE

Dataset Details

Dataset Description

This dataset includes antiviral assay results for dengue virus (DENV) infections across multiple time points (e.g., 12, 24, 36 hours post-infection) as well as dose-response data for selected hits. The dataset is structured to allow for time-series analysis of drug efficacy, and it is specifically designed to assist with the identification of potential antiviral compounds for DENV treatment.

• Number of samples: 4000+ images

• Time points included for time-series data: 0, 12, 24, 36, 48 hours post-infection (hpi)

• Columns in the dataset: CellProfiler features (intensity, sizeshape, texture), metadata, condition

• Acknowledgements: The author compiled this dataset based on work done by researchers in the Sexton, O'Meara and Tai labs at the University of Michigan.

Feature Extraction

High-content features were extracted using CellProfiler. CellProfiler is a powerful open-source tool for quantifying phenotypic changes from microscopy images. CellProfiler

Batch Correction with Harmony

To account for batch and plate effects, we used Harmony to correct the data at the image level. After Harmony correction, relevant metadata columns were appended back onto the dataset to allow for downstream analyses. Although this workflow used image-level data, it is fully compatible with cell-level or well-level data as well. Harmony

Dataset Sources

• Repository: Hugging Face Repository Link
• Noteboook Machine Learning Workflow
• Data Harmony Corrected Data is Managed by Git LFS
• Paper: Citation for DENV Assay Dataset Source

Disclaimer

Note: H2O's native XGBoost was not used in this notebook due to incompatibility with Windows (the OS used for this analysis).

Quickstart Usage

Install HuggingFace Datasets package

Each subset can be loaded into Python using the HuggingFace datasets library. First, from the command line install the datasets library:

$ pip install datasets

Optionally set the cache directory, e.g.,

$ HF_HOME=${HOME}/.cache/huggingface/
$ export HF_HOME

Then, from within Python, load the datasets library:

>>> import datasets

Load model datasets

To load one of the DENV-assay-TS-DR model datasets, use datasets.load_dataset(...):

>>> dataset_tag = "harmony_img_level.csv"
>>> dataset_model = datasets.load_dataset(
  path = "<HF PATH TO DATASET>",
  name = f"{dataset_tag}_models",
  data_dir = f"{dataset_tag}_models")['train']

The dataset is loaded as a datasets.arrow_dataset.Dataset:

>>> dataset_models
<RESULT OF LOADING DATASET MODEL>

Which is a column-oriented format that can be accessed directly, converted into a pandas.DataFrame, or parquet format, e.g.,

>>> dataset_models.data.column('<COLUMN NAME IN DATASET>')
>>> dataset_models.to_pandas()
>>> dataset_models.to_parquet("dataset.parquet")

Note: H2O's native XGBoost was not used in this notebook due to incompatibility with Windows (the OS used for this analysis).

Example Usage

You can filter the dataset by specific time points to analyze the drug response at different hours post-infection (e.g., 12 hpi, 24 hpi, etc.). The dataset includes features extracted from high-content cell images including intensity, sizeshape, texture and other related measurements.

Uses

This dataset is intended for researchers working on antiviral drug discovery, specifically those focusing on dengue virus (DENV). It is suitable for training models to predict antiviral efficacy based on drug dose-response data and time-series analysis.

Out-of-Scope Use

This dataset is not intended for use in human clinical trials or for making clinical treatment recommendations. It is intended solely for academic and experimental purposes.

Source Data

The source data for this dataset includes results from in-vitro DENV antiviral assays, with viral load measurements taken at multiple time points after infection. The dose-response assays were performed using various antiviral compounds to assess their effect on DENV replication.

Citation

@dataset{sfboisvert2025,
  title={DENV-assay-TS-DR Dataset},
  author={Sam Boisvert},
  year={2025},
  publisher={Hugging Face Datasets},
  howpublished={\url{https://huggingface.co/datasets/sfboisvert/DENV-assay-TS-DR}},
  doi={10.1234/dengue_assay_paper}
}