SPARCS combines AI microscopy with cell recovery to investigate gene function

Researchers at Ludwig Maximilian University of Munich and partner institutes described SPARCS on October 8, a method that uses AI and microscopy to examine genetically modified cells, then recovers selected cells for protein analysis. Their reported experiments examined images of 70 million cells, offering researchers a way to connect visible cellular changes with the genes behind them. The work was published in Cell, according to the university's report.

The collaboration was led by Veit Hornung at LMU's Gene Center, Matthias Mann at the Max Planck Institute of Biochemistry in Martinsried, and Fabian Theis at Helmholtz Munich. Their method screens images for complex visual characteristics after genetic changes have been made. Researchers can then isolate individual cells with features they want to investigate further.

How SPARCS links cell images to molecular changes

The recovery step is central to the approach described by the team. Once cells of interest have been selected through microscopy, the researchers isolate them intact and analyze their protein composition using mass spectrometry. That gives them a molecular readout to examine alongside the image and the genetic change. The university's account does not establish whether the recovered cells remain viable.

The 70 million figure refers to cells whose images were analyzed across the research, rather than to people tested or distinct genes identified. The published announcement describes two areas of investigation: autophagy, the process through which cells recycle material, and STING, a sensor involved in the innate immune system. Both provide examples of how a visual screen can direct researchers toward specific cellular mechanisms.

What the autophagy and STING experiments found

In the STING work, the researchers reported that the acidity of the Golgi apparatus, a structure that processes and moves material within cells, matters for STING's early transport. They linked the protein GPHR to that acidity and to the sensor's subsequent activation. These are findings about a cellular pathway; the announcement does not report a treatment effect or a clinical test.

How earlier image-based gene screening compares

SPARCS enters a field in which researchers have already used large-scale cell images to map gene function. In a separate study published in Nature Methods in January 2025, researchers associated with the Broad Institute and Calico reported a platform called PERISCOPE. Its atlas comprised three genome-wide maps built from gene knockouts across more than 30 million cells and more than 20,000 genes.

PERISCOPE combines staining that captures cellular features, optical sequencing of barcodes that identify genetic changes, and an analysis pipeline. Its authors reported grouping genes by their effects, reconstructing known pathways and protein interactions, and identifying TMEM251/LYSET as a Golgi protein needed to move lysosomal enzymes. This is an independent example of microscopy helping reveal intracellular processes; its results do not verify the newer SPARCS findings.

The distinction in the SPARCS account is the ability to select and recover individual cells after screening their images, allowing protein analysis of those cells. The studies used different designs, so their cell totals alone cannot show which method is faster, more accurate or more economical. Broad describes its earlier atlas as work in cultured human lung and cervical cancer cells, underscoring that these are laboratory research methods.

What remains to be established about SPARCS

The university identifies the new Cell paper as ‘SPARCS enables scalable recovery of complex image-based phenotypes for genetic screening,’ by Schmacke and colleagues. The team's account presents broader investigation of genes and biological mechanisms as the opportunity ahead. It supplies no timetable for wider use and no demonstrated patient benefit. Detailed measures such as cell-recovery rates and reproducibility cannot be assessed from the announcement alone.