A note from our CPO
Your archive can help pay for AI
I run product at Biobase. Every clinical organization we interface with is working out how to pay for AI. Part of the answer is in your archive. A Biobase integration lets you commercialize data, slide images and biospecimens, all while feeding it into a platform for your own AI agents. Security, compliance, and AI governance are all built in, and you control what's shared. Want to fund your AI deployment through the same platform you'll run your own AI stack on top of? Book time with me.
Chief Product Officer, Biobase
The Biobase Brief, issue 02 · October 2026. Adapted for the web from the Healthcare Partners edition.
Archived tissue at single-cell resolution
A comparison of 12 matched FFPE and fresh-frozen biopsies used a probe-based single-cell RNA sequencing platform. FFPE detected fewer genes per cell but assigned cell types with greater confidence and resolved immunotherapy-related biomarkers. At the same sequencing depth, the FFPE analysis also detected a subclonal MYC amplification missed in the fresh-frozen analysis.
The cohort was small. The result supports further evaluation of archival material for these workflows; it does not make FFPE universally preferable to fresh-frozen tissue.
Yan et al., Frontiers in Genetics 17, 1755978 (2026).
What a slide-based mutation model demonstrates
A Vision Transformer trained on more than 11,000 TCGA tumor samples across 32 cancer types predicted TP53 mutation status with a reported AUROC of 0.766. Survival prediction was less accurate on external validation.
Predicting mutation status from an image is a research signal. An AUROC alone does not establish equivalence to genomic testing, and the authors position the approach as complementary to sequencing.
Chaurasia et al., American Journal of Pathology (2026).
Block age is one attribute, not the whole assessment
Dong and colleagues compared Visium, GeoMx DSP, and Chromium FLEX using 14 archival tumors. Blocks were 22–103 months old, with DV200 values from 7.2% to 80.3%. Within this cohort, age and DV200 did not have a major effect on gene detection. The lowest-DV200 sample produced comparable results to other samples on Chromium and Visium.
The study does not make storage or RNA quality irrelevant to every assay. Its practical value is a reason to assess older material against a specific workflow rather than exclude it on age alone.
Dong et al., Nature Communications 16, 4400 (2025).
Give a potential partner enough context
Keep older collections visible in an initial inventory. Include collection dates, specimen format, linked pathology information, and known preparation or storage conditions. Mark missing information explicitly.
Then identify who can review research eligibility and availability. The collection preparation guide provides a starting structure for that conversation.