AI Could Transform Cancer Treatment by Precisely Identifying Tumor Cells

Heterogeneity Complicates Cancer Treatment

The international research team based their work on the fact that tumors are not composed of a single type of cell, but rather a combination of different cells. These cell types vary in how they proliferate and respond to therapy. Tumor microenvironment heterogeneity complicates the treatment of malignant neoplasms, and even the most advanced therapies may not achieve optimal results—triple-negative breast cancer being a prime example.

Associate Professor Christine Chaffer, study co-author and co-director of the Cancer Plasticity and Dormancy Program at Garvan, noted that tumors are still often treated as if they consist of just one cell type. A therapy targeting a specific mechanism of malignancy may eliminate only the subset of cells expressing that mechanism—leaving others behind to later proliferate and cause relapse.

Five Distinct Cell Populations

The results achieved with AAnet were recently published in Cancer Discovery. AAnet captures the functional state of tumor cells from multimodal data, detecting biological patterns that are highly relevant to clinicians. Until now, researchers were unable to precisely explain how cells within the same tumor differ, or how to tailor therapy accordingly, according to Associate Prof. Chaffer.

The researchers used AAnet to detect gene expression patterns in individual tumor cells, focusing on various models of triple-negative breast cancer.

The new AI tool identified five distinct groups of tumor cells in these models, each with unique gene expression profiles, suggesting significant variation in cell behavior. Each group exhibited different biological pathways, varying tendencies toward proliferation and metastatic spread, and markers associated with poor prognosis.

According to A/Prof. Chaffer, future research will aim to determine how these different cell types evolve over time and how they are affected by chemotherapy and other treatments.

Toward More Precise Therapies

AAnet was developed under the leadership of Associate Professor Smita Krishnaswamy from Yale University. She explains that detailed data on tumor cell subtypes allow for better analysis of their diversity and potential links to spatial tumor growth and metabolomic signatures.

Using AAnet to characterize tumor cell populations could, in A/Prof. Chaffer’s view, change the way cancer is treated. Currently, therapy decisions are primarily based on the affected organ and on specific molecular markers—assuming that all cells in the tumor behave similarly.

But AAnet clearly demonstrates that this is not the case. With AI, it is now possible to distinguish the behavior of individual cell populations, improving future combination cancer therapies by better targeting each of these subtypes.

Researchers hope that integrating AAnet into oncology diagnostics will enable more personalized treatments targeting all tumor cell types in an individual patient's cancer. Although the current study focused on triple-negative breast cancer, the authors believe that AAnet could be applied to other cancer types—and even to other diseases.

Editorial Team, Medscope.pro

Sources:

1. Venkat A., Youlten S. E., San Juan B. P. et al. AAnet resolves a continuum of spatially-localized cell states to unveil intratumoral heterogeneity. Cancer Discov 2025 Jun 24, doi: 10.1158/2159-8290.CD-24-0684.

2. Palma G., Frasci G., Chirico A. et al. Triple negative breast cancer: looking for the missing link between biology and treatments. Oncotarget 2015 Aug 30; 6 (29): 26560–26574, doi: 10.18632/oncotarget.5306.