AI Reads Gene Activity From Tissue Images
AI TIMES ·
✦ AI Summary
According to AI TIMES, a GIST research team has developed an AI framework that learns tissue images and gene expression data together, enabl…
According to AI TIMES, a GIST research team has developed an AI framework that learns tissue images and gene expression data together, enabling it to predict gene activity from images alone and identify areas inside tissue. The model was trained on about 980,000 pairs of tissue images and gene information, and the results were published in an international academic journal on September 28. The key is that it was designed not merely to match tissue appearance with gene activity separately, but to learn how the two types of information are connected. The research team said this suggests the model could be used for a broader range of analytical tasks than existing models tied to specific analyses. In actual tests, the team examined both its performance in predicting gene expression from tissue images and its ability to divide different regions within tissue. In particular, it showed that it could more precisely capture hidden boundaries and differences in characteristics in cases with complex internal variation, such as cancer tissue and brain tissue.
Perspective
This achievement can be read as an attempt to bring together analyses that have long operated separately: one that looks at tissue structure and another that reads gene activity. If a foundation model trained once can be extended to multiple analyses, researchers may face less burden in swapping tools, and the way subtle differences in diseased tissue are interpreted could become more sophisticated. Ultimately, the important shift is not competition over performance itself, but whether a common foundation for handling tissue data can be built.
This perspective is BizCrush's own commentary and is not part of the reporting by AI TIMES.
This article was produced with the help of an automated content generation algorithm.
Source: AI TIMES
View originalThis article was summarized and organized by BizCrush based on the original article from AI TIMES. For exact quotations and full details, please refer to the original article.