ELEM Biotech plans to invest US$1 million in publicly shared perturbation data for virtual cell model training and validation. We invite researchers worldwide to tell us which experiments would make that data useful.

01 / Start with the experiments you need

A useful virtual cell dataset begins with a concrete question: what happens when a particular gene is perturbed in a particular cell, under defined conditions?

Our next dataset should reflect those questions—not simply the samples we can produce.

Based on the already released open-source bulk RNA-seq dataset ELEM-1200RS (https://huggingface.co/datasets/ELEM-Bio/ELEM-1200RS) and the upcoming open-source proteomics & RNA-seq dataset ELEM-600PR, which contains more than 1,200 bulk RNA-seq and 600 proteomics samples across multiple human cell backgrounds, we are seeking specific experimental priorities from academic groups, companies, open-source teams and independent researchers.

Which cells? Which genes? Which perturbations?

Tell us whether you need knockout, knockdown, activation, or combined gene–drug conditions. Identify the sampling times, readouts, controls and independent biological replicates that would make the results interpretable.

A ranked list of concrete cell–gene–perturbation combinations is more useful than a broad wish list. Explain which experiments are essential, which are optional, and how the data would support a real biological or modeling task.

02 / Scalable experiments, carefully chosen measurements

ELEM’s automated platform will prepare up to produce up to 100,000 arrayed genetically perturbed cell samples per month . It supports combined genetic and drug perturbations and sampling at multiple time points.

Samples are compatible with single-cell RNA sequencing, bulk RNA sequencing, quantitative proteomics and cell imaging.

This is cell-sample production capacity—not a promise of 100,000 distinct genes or fully measured multimodal datasets each month. Final coverage will depend on the chosen cells, perturbation feasibility, controls, replication, quality requirements and measurement costs. The final design will specify the selected readouts and how measurements are paired.

The goal is not simply more samples. It is data that researchers can interpret and models can be tested against.

03 / A community-informed design—not a popularity contest

ELEM plans to allocate US$1 million to data generation . This call seeks scientific input; it is not a donation request or a cash-grant competition.

Requests will be assessed for scientific value, community relevance, gaps in existing public data, feasibility and cost. Submission does not guarantee selection, and financial contributions will not purchase priority or exclusive access.

We propose a 60-day public comment period on the draft design and publication of the final experimental design within 30 days of program launch. The design will identify cells, genes, interventions, time points, controls, independent biological replicates, readouts, quality criteria, budget allocation and the evaluation protocol.

04 / Open by design, with a public release schedule

For the new dataset, our intended policy is free, non-exclusive worldwide access under CC BY 4.0, including commercial use , without an application gate or preferential access for contributors.

We propose a first public release within six months of program launch, and release of all completed, shareable data within twelve months. Failed and excluded experiments will be documented. Any data temporarily withheld for blinded model evaluation will have an advance-announced release date and become public by the same final deadline.

Planned releases include raw and processed measurements, sample-level metadata, quality-control results and protocols, together with separately licensed open-source analysis code. Privacy and third-party rights safeguards will apply.

Funding approval, scope, exact release dates and license terms will be confirmed publicly before experiments begin. These proposed terms do not change the existing ELEM-1200RS license. Payment, participation and an email subscription will not be required to access the new public data.

05 / Help define the next experiments

Submit your priorities by December 25, 2026, at 23:59 UTC.

If you received our invitation email, complete the attached 19-question survey and reply with your responses. Where possible, include the optional experiment-request template. Alternatively, reply with your top three experimental priorities, and we will follow up.

If you are joining through this article, email [yongbo_cheng@elem-bio.cn] to request the questionnaire and template, then return your completed responses by the same deadline.

You do not need to share model weights, unpublished results or proprietary methods. Submit only information you are authorized to share. Your contact details will not be included in the public experimental dataset, and future program emails are optional.

Researchers interested in reviewing the design, contributing protocols or participating in model validation are also welcome to indicate their interest.

Tell us what is missing. Help us decide what to measure next.

ELEM Biotech

ELEM Biotech develops genetic perturbation tools, engineered cell models and experimental data resources for biological research and AIVC model. Its automated arrayed perturbation platform supports genetic and combined gene–drug experiments, time-resolved sampling and multimodal data collection.

Experimental data for biological understanding and virtual cell model development.

Existing dataset: https://huggingface.co/datasets/ELEM-Bio/ELEM-1200RS

Program details and public updates: https://huggingface.co/datasets/ELEM-Bio/Open_Virtual_Cell_Data_Initiative.

website: https://en.elem-bio.cn/