What if the key to treating liver and lung diseases were hidden in a medicine that has been present in millions of households for decades? With the support of the Early Adopters BIO+IA programme, a team of researchers from Colombia and Venezuela used artificial intelligence and molecular network analysis to investigate a possibility that could open up new horizons for the medicine of the future. The result? A demonstration of the value of regional collaboration and of how artificial intelligence can accelerate scientific discovery in Latin America and the Caribbean.
What if a medicine used for over 60 years to relieve coughs had the potential to help with serious diseases for which there are currently few treatment options? That was the question a binational team, comprising researchers from the Autonomous University of Bucaramanga (Colombia, an active member of the national academic network, RENATA) and the Venezuelan Institute for Scientific Research (IVIC), decided to address as part of the Early Adopters BIO+IA programme – an initiative promoted by RedCLARA within BELLA II to bring advanced bioinformatics and artificial intelligence capabilities to researchers in Latin America and the Caribbean.
The scientific group’s aim was to study the potential of dextromethorphan, a well-known cough suppressant, as a possible treatment for fibrotic diseases, particularly those affecting the liver and lungs. To this end, the team utilised artificial intelligence tools, molecular network analysis and regional scientific collaboration.
A complex question requiring new tools
Fibrotic diseases represent a major health problem. They are chronic conditions in which an organ suffers damage or persistent inflammation, leading to an excessive build-up of scar tissue (extracellular matrix) that replaces healthy cells. Over time, this causes the organ to become stiff, reduces its elasticity and progressively destroys its structure and biological function. This occurs, for example, in pulmonary fibrosis and various liver diseases.
Not suitable for children under 2 years of age, dextromethorphan is a cough suppressant that soothes a dry or irritative cough caused by colds or flu; available over the counter in syrup and tablet form, this chemical compound is widely and commonly used because, although it does not cure the underlying illness nor is it effective for a productive cough, it does provide relief by acting on the brain to suppress the cough reflex.
Studies carried out since 2023 have shown promising signs regarding the anti-inflammatory and anti-fibrotic effects of the very same dextromethorphan used to treat coughs in fibrotic diseases; however, there was still no clear explanation as to how the medicine might act on the biological mechanisms involved in these processes.
The challenge was enormous: to analyse vast amounts of scientific information scattered across thousands of articles, identify connections between genes, proteins and biological processes, and construct a comprehensive picture capable of transforming isolated clues into a sound hypothesis.
The value of the Early Adopters BIO+IA programme
The project was one of those selected to participate in the Early Adopters BIO+IA programme, an initiative created to enable research teams from the region to work with the BELLA II Bioinformatics and Artificial Intelligence Testbed, an infrastructure designed to capture, process and analyse scientific information using generative artificial intelligence, logical artificial intelligence and structural bioinformatics.
According to the researchers, the main motivation for applying was that their hypothesis was “biologically sound but computationally unfeasible with local resources”. The researchers Gustavo Antonio Bruges Morales (Autonomous University of Bucaramanga), Rafael José Puche Quiñonez (IVIC) and Fernando Antonio Hernández Medina (Venezuelan Institute for Scientific Research), who led the study, maintain, in a written interview, that “the hypothesis as to whether dextromethorphan, a cough suppressant with over sixty years of clinical use, can be repurposed as a multi-organ antifibrotic cannot be resolved by a single trial or a conventional literature search: it requires the construction and analysis of complete molecular networks, integrating hepatic and pulmonary transcriptomics, protein-protein interactions and the mining of tens of thousands of abstracts”.
Access to the testbed provided them with the technological infrastructure, specialised tools and technical support needed to carry out analyses that would have been extremely difficult to perform independently.
“This type of work has three bottlenecks that a regional university would struggle to resolve on its own: persistent computing infrastructure (our inference runs generate trace files of over 140 MB and take hours), access to specialised tools that are not packaged for straightforward installation, and —most difficult of all— technical support from those who developed those tools. The BIO+IA testbed offered all three in one place,” explain the specialists.
When artificial intelligence helps to challenge one’s own hypotheses
One of the greatest contributions of the experience was that the technology was not used to confirm a preconceived idea, but rather to put it to the test.
By analysing thousands of scientific publications and constructing molecular interaction networks, the researchers observed that some of the mechanisms they initially considered central lacked sufficient scientific support, whilst others emerged as far more significant factors than expected.
The platform enabled the processing of tens of thousands of scientific records and the construction of knowledge bases containing thousands of documented biological relationships, whilst always maintaining traceability back to the original sources.
For the team, this was one of the most valuable lessons learnt: “The most important thing was that the testbed forced us to turn a narrative hypothesis into a falsifiable structure, and then gave us the tools to refute ourselves”.
As the analyses progressed, a new understanding of the problem emerged. What began as a hypothesis based on a simple relationship between a drug and a biological target evolved into a much broader vision based on networks of molecular interactions.
Regional collaboration as a key element
Beyond the scientific results, this project demonstrates the value of the collaboration promoted by RedCLARA, the BELLA II project it implements, and the Latin American National Research and Education Networks (NRENs) that bring it to life.
The binational research team was supported by RENATA, the Colombian RNIE, working on a shared regional infrastructure. According to the participants, this capacity for collaboration was a fundamental part of the experience: “The programme consolidated a Colombia–Venezuela collaboration operating on a shared infrastructure. For groups in the region, that access is the difference between posing a systems biology question and being able to answer it.”
This coordination between researchers, national networks and regional platforms is precisely one of the central objectives of the BELLA II project: to enable knowledge, infrastructure and talent to come together to tackle scientific challenges that transcend borders. Co-funded by the European Union, this type of contribution – where cooperation within the ecosystem of the RNIE networks that make up RedCLARA generates results that benefit people – forms the very foundation of the BELLA II project.
Results that go beyond a single project
During their participation in Early Adopters BIO+IA, the team succeeded in building a reproducible mechanistic model, identifying new lines of research and developing capabilities that will remain in place at their institutions.
They also consolidated reproducible tools in R and Python, strengthened their capabilities in biological network analysis and established open repositories aligned with the principles of open science.
However, for the researchers, the main achievement was having raised the standard of evidence available to support their hypotheses. As the team summarises: “The testbed allowed us to move from a hypothesis that could be defended in a seminar to a set of claims traceable back to the original sentence and PMID, which is the standard required for a publishable manuscript”.
An example of BELLA II’s potential to accelerate regional research
This experience demonstrates how the combination of artificial intelligence, bioinformatics, shared infrastructure and international collaboration can significantly accelerate scientific research processes.
In this case, the BIO+IA Testbed not only helped to generate new knowledge about the therapeutic potential of dextromethorphan; it also enabled researchers from different countries to work together, challenge their own hypotheses and build a more robust evidence base for future research.
This is precisely the kind of impact that RedCLARA and BELLA II are seeking: to transform connectivity into effective collaboration and turn access to advanced technologies into real opportunities for Latin American science. This is, moreover, a concrete example of the human-centred approach of the EU-LAC Digital Alliance, of which BELLA II is a cornerstone.
