A research team from the Pia Cosma Lab at Guangdong Provincial People’s Hospital has collaborated to develop an artificial intelligence algorithm called AINU (Artificial Intelligence of the NUcleus), which can identify cell types, distinguish between cancerous and normal cells, and detect the very early stages of viral infections within cells. The findings were published in Nature Machine Intelligence (Zone 1 Journal of the Chinese Academy of Sciences) on August 27, 2024, under the title “A Deep Learning Method that Identifies Cellular Heterogeneity Using Nanoscale Nuclear Features”. These discoveries pave the way for using AI technology to improve early cancer diagnosis, identify potentially metastatic cancer cells, or diagnose viral infections in clinical settings.

Compared to normal cells, cancer cells or virus-infected cells exhibit significant changes in nuclear structure, such as altered DNA organization or changes in the distribution of enzymes within the cell nucleus. When cells are infected by a virus, the tightness of DNA packaging undergoes dramatic changes. These early changes are extremely subtle and are difficult to detect using traditional methods at an early stage. “Super-resolution images” are obtained using stochastic optical reconstruction microscopy (STORM), which can display the nuclear structure of cells at the nanoscale level, allowing for higher resolution details than conventional microscopes and revealing details as small as 20nm (5,000 times smaller than the width of a human hair).

Figure 1: Dual-Color STORM Imaging of Induced Pluripotent Stem Cells and Somatic Cells

AINU can classify cell types by scanning super-resolution images that reveal cell structures at the nanoscale. After training, the algorithm can categorize cells as cancerous or normal based on these features. The nanoscale resolution of super-resolution microscopy allows AI to detect nuclear changes in cells infected with type 1 herpes simplex virus (HSV-1) within an hour. The model can detect the presence of viruses by identifying subtle differences in DNA packaging tightness. Researchers can use this technology to understand how viruses affect cells upon entering the human body, which may help in developing better treatment strategies. Clinically, AINU can be used to quickly and accurately diagnose early-stage cancer and viral infections from simple blood or tissue samples.

This research was a collaborative effort between the Pia Cosma Lab at Guangdong Provincial People’s Hospital, the Centre for Genomic Regulation, Universitat Pompeu Fabra, and the University of the Basque Country. The first authors are Zhong Limei from the Pia Cosma Lab, Department of Medical Research and Davide Carnevali, a postdoctoral researcher at the Centre for Genomic Regulation. Ignacio Arganda-Carreras from the Computer Science and Artificial Intelligence Center, University of the Basque Country provided valuable assistance in the algorithm’s development.

Zhong Limei

Updated: 10 September, 2024