Rivercell's $25 million virtual-cell plan starts with experimental data

Rivercell has announced a $25 million seed round to expand the experimental foundation for an AI “virtual cell”: a model intended to predict how human cells respond to drugs or genetic changes. For biotech research teams, the development is investment in data generation and model building. It is not evidence that the company has improved a treatment.

The Paris company's October 7 announcement, distributed by Business Wire, names HV as lead investor, alongside HCVC, Alven and Bpifrance Digital Venture. The funds are intended to scale its data platform, expand its Paris wet lab and launch its AI Virtual Cell program.

The laboratory is part of the model-building process

Rivercell describes its approach as generating interventional, time-resolved and multimodal single-cell data. In practical terms, those labels mean applying a change, observing responses over time and collecting different types of measurements from individual cells. It is developing purpose-built laboratory infrastructure to produce the training data.

That design addresses a specific problem: a snapshot showing two features together does not necessarily explain what will happen if a researcher changes one of them. In an October 1 account of its own virtual-cell work, Arc Institute explains why controlled perturbation data matter. Its Perturb-seq experiments pair a targeted genetic intervention with the resulting gene-expression profile.

Rivercell's proposed data engine and Arc's experiments are separate efforts. Their shared relevance is the need to connect an intervention with a measured response. More computing cannot substitute for knowing which intervention occurred or for reliable measurements of its consequences.

A useful prediction must travel beyond familiar examples

Arc's 2026 Virtual Cell Challenge tests predictions across six cell lines, emphasizing biological contexts for which a model has not been shown perturbation responses. This provides a concrete example of a demanding evaluation question: does a method work when the cellular setting changes?

It does not establish Rivercell's performance. A research partner would still need to see how Rivercell separates training and evaluation data, which cell types and interventions it covers, and how predictions compare with appropriate baseline methods. Results should make clear where the model fails as well as where it succeeds.

The company's ambition is to reduce the need for physical experiments by predicting some outcomes computationally. The useful near-term evidence would be prospective predictions followed by laboratory checks, with enough methodological detail to assess them. A funding announcement supplies resources for that work; the next substantive milestone is a documented prediction that helps researchers choose or interpret an experiment.

Sources

• Rivercell announcement distributed by Business Wire, October 7.

• Arc Institute: data generation for the 2026 Virtual Cell Challenge, October 1.

Research note: Company plans are attributed to Rivercell's release. Arc's documentation supplies field context, not a test of Rivercell. TechKili has not independently evaluated the platform or any clinical outcome.