Forecast Bio discloses $2.4M SAFE financing for its astrocyte AI platform

Why it matters

Forecast is testing whether astrocyte activity can become a predictive measurement for psychiatric drug development, giving developers another way to assess candidates before clinical trials.

Forecast Bio, founded by UCSF neuroscientist Kira Poskanzer, has disclosed $2.4 million in SAFE financing for an AI platform intended to predict how psychiatric drugs affect human brain circuits and symptoms.

Forecast Bio reported the financing in a Form D filed with the SEC on August 27. Forecast has sold $2.4 million of a targeted $3.15 million offering to nine investors, leaving $750,000 available. The filing does not identify the investors, a lead backer, the valuation or the conversion terms attached to the Simple Agreements for Future Equity.

The date matters. Forecast's filing is new, while the financing began on November 11, 2025, according to the Form D. The disclosure therefore captures an offering that has been underway for about nine months. Forecast incorporated in Delaware in 2025, and this is its first listed Form D under the SEC identifier used for the filing.

The filing identifies Poskanzer as an executive officer and director. Her route into biotech entrepreneurship runs directly through the type of biology Forecast is trying to commercialize. She is an associate professor of biochemistry and biophysics at the University of California, San Francisco, where her laboratory studies interactions between neurons and astrocytes using two-photon imaging, electrophysiology and tools for controlling cellular activity.

Poskanzer earned a biology degree from Brown University, a neuroscience PhD from UCSF and completed postdoctoral training at Columbia University. Her published work has examined how astrocytes respond to neurotransmitters and help regulate cortical state and sleep. Forecast takes that academic focus and turns it toward one of drug development's harder problems: determining whether preclinical results will predict what happens in patients.

Turning astrocytes into a drug-screening signal

Forecast is building the company with co-founder Maxine Levesque, a physician-scientist whose background combines neuroscience, biomedical engineering, mathematics and software. Levesque studied biomedical engineering and mathematics at Johns Hopkins University before entering UCSF's medical scientist training and neuroscience programs. Her research has included astrocyte calcium activity, intracranial brain recordings and the use of language models to extract information about changes in mental state.

That combination explains Forecast's technical premise. Astrocytes, once treated largely as support cells for neurons, form signaling networks that respond to chemical and physiological changes in the brain. Poskanzer and Levesque are betting that those responses contain enough structured information to help forecast the effects of drug candidates.

The founders offered an early view of that work in OpenAstrocytes, an open repository of nearly 1 million images of astrocyte calcium dynamics from mouse cortical experiments.

Forecast says computer-vision embeddings can identify anatomical and physiological features in the images and decode responses to compounds applied to the tissue.

The published work gives Forecast a data and analysis starting point. Forecast's larger commercial thesis requires several additional steps: moving from mouse brain slices to human-derived systems, producing repeatable signals across experiments and showing that those cellular measurements predict patient symptoms or adverse effects closely enough to influence drug-development decisions.

Forecast's published roadmap points toward three-dimensional systems grown from human induced pluripotent stem cells, along with neuronal datasets that complement its astrocyte work. The intended product would use AI to connect activity measured in those laboratory systems with outcomes observed in people.

A preclinical bet on psychiatric drugs

Forecast describes its target as a single preclinical assay capable of mapping how a prospective drug affects brain circuitry, symptoms and side effects. One Mind selected Forecast for its 2026 accelerator cohort and described the platform as potentially speeding therapeutic development by 10 times. That figure remains a program objective rather than a demonstrated clinical or commercial result.

The appeal is straightforward. Central nervous system drug programs are expensive, lengthy and vulnerable to failures when findings from animal models or simplified cell cultures do not translate into human outcomes. A laboratory assay with meaningful predictive power could help developers reject weak candidates earlier and decide which compounds warrant further investment.

Predictive power is also the central risk. Human-derived brain models can vary between batches and may lack the maturity and biological complexity of a living brain. Forecast will need to show that its AI is finding signals tied to treatment outcomes, rather than patterns peculiar to a dataset, imaging process or laboratory system.

The financing gives Poskanzer and Levesque room to pursue that validation. Forecast has already secured institutional footholds beyond the One Mind program. Merck selected Forecast for its 2026 Digital Sciences Studio, which focuses on technology for drug discovery and development. Merck said each selected company would receive $100,000 through a SAFE funded by the Merck Global Health Innovation Fund, The Venture Collective and CQDM.

The public records do not establish whether that $100,000 commitment is included in Forecast's $2.4 million Form D total, so it should not be added as separate funding. The SEC filing establishes that at least $2.4 million has been sold in the current offering. It leaves the identities of Forecast's nine investors and the price of their bet undisclosed.

For Poskanzer, Forecast is an attempt to carry years of astrocyte research out of the laboratory and into the decisions that determine which brain drugs reach patients. Forecast's open dataset shows where the founders are starting. The financing is paying for the harder part: proving that cells in a dish can offer a useful forecast of the human mind.