A note from our CEO
Welcome to the Brief
Biobase exists because clinical data is abundant and still remarkably hard to use. Most of what slows research down isn’t the science, it’s friction: finding the right material, knowing whether it’s usable, then cleaning, normalizing and moving it without months of back and forth.
This newsletter is what we see working on that problem. Once a month, briefly: new methods, clinical developments, and what they mean for the people sourcing tissue and those supplying it.
If any of this touches what you’re working on, I’d like to hear about it.
Chief Executive Officer, Biobase
The Biobase Brief, issue 01 · September 2026. Adapted for the web from the Healthcare Partners edition.
A proteome from one slide
Berkovska and colleagues profiled FFPE lung cancer biopsies from 68 patients using a single 4-micron section per patient. The analyses identified up to approximately 8,000 proteins per sample, including drug targets and immune-related signals.
This supports retrospective research on limited material. It does not promise the same depth from every slide: tissue composition, handling, and the analytical workflow still matter. Pathology context helps make the resulting measurements interpretable.
Berkovska et al., npj Precision Oncology 10, 196 (2026).
Match the method to the tissue
A comparison of three commercial spatial transcriptomics platforms used 33 archival tissue types without pre-screening for RNA quality. Xenium showed the highest per-gene sensitivity; MERSCOPE was particularly sensitive to input quality. This is a comparison under the study’s conditions, rather than a universal ranking of platforms.
Wang et al., Nature Communications 16, 10215 (2025).
A separate method from the Lau lab recovered intact cells from FFPE colon tissue, rather than isolating nuclei alone. RNA recovery and representation of some cell types improved in colon, while gains in thymus were limited. The tissue type remains part of the method decision.
Evans et al., Cellular and Molecular Gastroenterology and Hepatology (2026).
The signature in your slides
A benchmark across 34 medical centers tested 20 pathology foundation models. All encoded information about the slide’s source center. Laboratory procedures, staining, and scanner differences can be learned alongside biological features, with consequences for downstream performance.
This gives research teams a reason to ask about source institutions and imaging conditions when assembling or evaluating a dataset. It does not establish a commercial value for any particular collection.
Kömen et al., Nature Communications (2026).
Make the collection assessable
A useful inventory makes preparation, linked pathology context, and imaging provenance visible. Record what is known, identify gaps, and name the people who can assess availability and proposed research use.
Start with the collection preparation guide. A clear description gives a prospective partner something concrete to evaluate without implying that material is already approved for release.