A microscopic image of tissue is like a city map: streets, houses, parks. An experienced eye can tell a lot about a neighborhood from such a map. But what if you could predict from the map which houses are noisy today and which are quiet? In biology, a similar challenge exists: from a routine tissue slide stained with standard dyes (hematoxylin and eosin), scientists want to know which genes are active at each point. Usually, this requires the expensive method of spatial transcriptomics, which gives a precise picture but is slow and costly. The new approach, CHRep, achieves the same from a routine slide — quickly and cheaply.

CHRep works in two stages. First, it learns to see tissue structure: how cells are arranged, where dense clusters are, where empty spaces are. This is like learning to read a map, remembering that parks are usually green and rivers are blue. Then, when the model is applied to slides from another laboratory (where lighting or staining may differ slightly), calibration kicks in. It resembles tuning an old radio: without changing the whole circuit, you turn one knob to get a clear signal. Thanks to this, CHRep is robust to differences between hospitals and laboratories.

Such a tool could accelerate studies of large patient groups and help doctors diagnose faster from routine biopsies. Instead of expensive and lengthy analysis, a simple slide and an accurate prediction.