<b>HyperMap: An Efficient Meta-Learning Framework for Transferring Perturbation Responses

A 15.8 GB resource of predicted transcriptional responses to 19,036 genetic and chemical perturbations across 19 human cell lines. Predictions were generated using the HyperMap meta-learning framework and contain single-cell gene expression deltas for 2,500 genes. The dataset was authored by Bhavya Dhaka and last updated on March 22, 2026.

Use Cases

• Benchmarking meta-learning models for biological knowledge transfer based on the described HyperMap framework.
• Analyzing predicted gene expression changes for 2,500 genes in response to specific chemical or genetic perturbations.
• Comparing predicted perturbation responses across the 19 different human cell lines included in the dataset.
• Training or fine-tuning models for in-silico perturbation experiments based on the predicted single-cell expression deltas.

Strengths

• Large scale of 19,036 distinct perturbations provides broad coverage of potential experimental conditions.
• Unified predictions across 19 human cell lines allow for comparative analysis.
• Predictions cover 2,500 genes, offering a detailed transcriptional profile.

Limitations

• Data consists of model predictions, not direct experimental measurements; accuracy depends on the underlying HyperMap framework.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Predictions generated using the HyperMap meta-learning framework, transferring knowledge from existing perturbation atlases.

Time Range

Predictions are not tied to a specific experimental date range.

Freshness

Last updated 2026-03-22 23:02:50; freshness should be verified.

Geography

null