Strand AI

Researchers working with archival tissue cohorts or retrospective trial samples often cannot answer spatial biomarker questions because the assay that would answer them, such as multiplex immunofluorescence, was never run and the tissue blocks were never intended for spatial work. Running real multiplex assays consumes precious archival sections, often making whole-cohort or retrospective studies impractical. Teams are left choosing a small shortlist of samples to profile, effectively guessing which ones justify the cost of wet lab work, while the rest of the cohort stays unexamined. The core issue the page describes is that most patient records are shallow: cohorts get built around whichever assays were affordable at the time, and the measurement that would resolve a question is usually the one nobody took.

Strand AI's product, Lattice, is a model that predicts spatially resolved protein-marker maps directly from standard H&E-stained slides, delivering the output as virtual multiplex immunofluorescence (mIF) channels at the resolution of the original image. Because the prediction is computational, no additional tissue is consumed and archival blocks remain intact, making it possible to run across whole retrospective cohorts rather than a handful of selected sections. The model is described as pan-cancer and pan-tissue, developed on thousands of tissue regions and millions of image patches spanning more than a hundred protein markers, and validated at pixel, cell, region and population levels against real stains and multiplex data. Strand AI runs this as an end-to-end dry lab service: partners send digitized H&E slides, Strand AI runs Lattice along with cell segmentation, phenotyping and tumor microenvironment characterization, and returns structured spatial data plus a written analysis targeted at the partner's biomarker question. For cohorts that cannot leave an institution, the company states it will deploy and run Lattice within that environment instead.