A Chilean collaboration driven by BELLA II: Artificial intelligence to understand resistance to triple-negative breast cancer

With the support of RedCLARA, REUNA and the BELLA II BIO+IA testbed, a Chilean team of researchers from the Universidad del Desarrollo (UDD) and the University of Chile (UCh) used artificial intelligence to identify new biological pathways associated with resistance to treatments for triple-negative breast cancer.

Between 10% and 15% of all breast cancer cases are triple-negative; this is one of the most aggressive subtypes of the disease. It typically grows and spreads faster than other types and carries a higher risk of relapse in the early years. Triple-negative breast cancer most commonly affects young women under the age of 40 and is sometimes linked to genetic mutations.

Its high mortality rate and the limited availability of effective treatments make drug resistance one of the main challenges for patients, doctors and researchers. For Patricia Silva Morales, a PhD student in Science and Innovation in Medicine at the Universidad del Desarrollo, understanding why some tumours stop responding to drugs such as cisplatin and olaparib was an urgent scientific question. The answer began to take shape thanks to a regional initiative.

In 2026, the project ‘The Role of Mitochondrial Candidates in the Drug Resistance of Triple-Negative Breast Cancer’, developed by the Universidad del Desarrollo in collaboration with the University of Chile and with the support of REUNA, was one of five selected to participate in the Early Adopters BIO+IA Programme, promoted by RedCLARA within the framework of BELLA II. The programme was created to enable research teams from Latin America and the Caribbean to access advanced artificial intelligence capabilities applied to bioinformatics via the BIO+IA Testbed.

Collaborative science to tackle a complex problem

The team comprised researchers Patricia Silva Morales, Ricardo Armisén (both from UDD) and Katherine Marcelain (Uch). This collaboration was joined, from the academic networking ecosystem, by REUNA, the Chilean National Research and Education Network, and RedCLARA, with the Bioinformatics and Artificial Intelligence (BIO+IA) testbed developed as part of the BELLA II project, which is co-funded by the European Union. The collaboration enabled the team to combine expertise in oncology, functional cancer genetics, bioinformatics and biological data analysis to tackle a common challenge: understanding how certain mechanisms associated with mitochondria allow tumour cells to survive treatments designed to destroy them.

The research was based on an innovative hypothesis: specific modifications to mitochondrial RNA (a set of ribonucleic acid (RNA) molecules produced within mitochondria through the transcription of their own DNA) could act as ‘biological switches’ capable of promoting the survival of tumour cells in the face of chemotherapy. To test this idea, the team needed to integrate large volumes of information from biomedical databases, scientific literature and molecular analyses.

The contribution of the BIO+IA testbed

BELLA II’s Bioinformatics and Artificial Intelligence testbed was designed precisely to tackle this type of challenge. The platform integrates generative and logical AI, along with bioinformatics tools capable of analysing scientific information scattered across numerous specialised repositories, facilitating the identification of molecular interaction networks and the generation of experimental hypotheses.

Over eight weeks of work, the participating teams received specialised training and technical support, utilising a regional infrastructure operated on RedCLARA and enabled by BELLA II. The aim was to accelerate research processes that normally require months or even years of literature review and manual analysis.

For Patricia Silva, access to these capabilities proved transformative: “We decided to apply to the Early Adopters Programme because we recognise the transformative value that Artificial Intelligence brings to current biomedical research. Specifically, we were looking to enhance the speed, depth and scope of our literature reviews and data analyses.”

The researcher explains that the main value of the testbed was its ability to significantly expand the scope of previously conducted bioinformatic analyses and to strengthen the hypotheses that are currently being validated experimentally.

New therapeutic hypotheses and concrete results

One of the main benefits of using AI was the ability to analyse large volumes of scientific literature and convert that information into comprehensible models of biological interaction. “What we consider to be a distinctive contribution was the ability to translate the vast knowledge contained in thousands of papers into visual diagrams representing these mechanisms,” states the UDD specialist.

This process enabled the research team to identify new molecular associations linked to treatment resistance and to discover biological pathways that had not been considered in the original research proposal.

Among the most significant findings was the identification of the central role played by reactive oxygen species (ROS) in the response to cisplatin, as well as the possible involvement of the CISD1 and ALDH4A1 genes in tumour survival mechanisms. What does this mean? It means that cisplatin damages tumour cells via molecules called ROS, but the CISD1 and ALDH4A1 genes could help the tumour defend itself and survive; understanding this process could lead to more effective cancer treatments in the future. This is a major finding of this research.

Furthermore, the analysis suggested a new therapeutic strategy based on drug repositioning. Artificial intelligence enabled the identification of evidence positioning the well-known drug pioglitazone as a potential adjunct to enhance the effectiveness of treatment in resistant cases.

For the team, the impact went beyond the specific findings. “Taking part in this programme represented a methodological turning point for our research,” says Patricia Silva, who explains that the research succeeded in moving from the analysis of individual genes towards an understanding of complete tumour survival networks, providing a broader view of the biological processes involved in drug resistance. Result: new therapeutic hypotheses were developed, ready to progress to experimental validation stages, and the team’s capabilities in the use of artificial intelligence applied to biomedical research were strengthened.

The value of a regional infrastructure

The outcome of the “Role of mitochondrial candidates in the drug resistance of triple-negative breast cancer” project, carried out using the BIO+IA testbed, reflects one of BELLA II’s central objectives: to enable researchers from Latin America and the Caribbean to access infrastructure, technological capabilities and regional collaboration to tackle high-impact scientific problems.

The BIO+IA Early Adopters programme is a concrete example of how cooperation between RedCLARA, the National Research and Education Networks and academic institutions can accelerate the generation of scientific knowledge for the benefit of the region. Proof of this is provided by Patricia Silva’s remarks: “The greatest achievement that made a real difference was obtaining predictive validation of our candidate genes and generating a mature therapeutic hypothesis ready for in vitro testing, thereby greatly optimising our resources and laboratory time.”

More than just a new technological tool, the BIO+IA Early Adopters Programme demonstrated that the combination of artificial intelligence, scientific collaboration and regional infrastructure can pave faster paths to understanding complex diseases and bring new therapeutic opportunities to those who need them most.