Malaga scientists use artificial intelligence to identify “backup genes” for serious diseases

A Malaga research team has developed an artificial intelligence method capable of predicting which genes could compensate for the loss of function of other disease-related genes.

The study opens a new route for identifying potential therapeutic targets in genetic disorders and understanding why certain tumours develop resistance to medicines.

The tool is not a treatment and cannot currently prevent a disease directly. Its purpose is to analyse large quantities of biomedical data and identify the most promising candidates for subsequent laboratory research.

Scientists from the University of Malaga, IBIMA Plataforma BIONAND, Spain’s Rare Diseases Biomedical Research Network and the Spanish National Bioinformatics Institute took part in the project.

What are “backup genes”?

The system is based on a natural phenomenon known as functional genetic compensation.

Under certain circumstances, when one gene stops working correctly, another gene with a similar activity can perform part of its function. This compensation may reduce the effects of the original genetic alteration on cells and the wider body.

Finding these alternative genes could make it possible to explore a strategy that does not depend exclusively on repairing the defective gene. Researchers could instead investigate ways to increase the activity of the gene providing the backup function.

This approach is particularly relevant to genetic diseases. However, testing every possible relationship among thousands of human genes through conventional experiments would be exceptionally difficult.

How artificial intelligence is used

The Malaga team combined bioinformatics, network analysis and machine learning to reduce the enormous number of possible gene combinations.

The method integrates information from eight major biomedical resources. The researchers used this data to construct 19 functional networks connecting thousands of genes.

The system analyses information including:

• Interactions between proteins.
• The biological functions of genes.
• Associations with diseases and symptoms.
• The cellular pathways in which genes participate.
• The genetic dependencies of tumour cells.
• Similarities between different genes.
• Existing information about genetic alterations.

These relationships are converted into mathematical representations using techniques known as graph embeddings. One of the methods, node2vec, can detect connections and similarities that may not be obvious through conventional analysis.

Reducing thousands of candidates to a manageable list

The tool does not automatically confirm that one gene can replace another. Instead, it ranks candidates according to their predicted probability of performing a compensatory function.

This prioritisation could save considerable time and resources. Laboratories would not need to test thousands of relationships indiscriminately and could focus on the genes ranked as the strongest candidates.

The method also uses NetAnalyzer, a biomedical-network processing tool previously developed by the research group and now expanded to incorporate new graph-learning techniques.

Results involving three genetic diseases

To assess its performance, the team tested the system using diseases where compensatory genes have already been scientifically demonstrated.

The known relationships appeared among the tool’s leading candidates in three conditions:

• Duchenne muscular dystrophy.
• Spinal muscular atrophy.
• Beta-thalassaemia.

This shows that the model can recover previously validated genetic connections and supports its use to suggest candidates for diseases where compensatory mechanisms have not yet been identified.

However, any newly predicted genes must undergo biological validation before researchers can consider a potential clinical application.

A possible route for tackling cancer drug resistance

Genetic compensation can also have harmful consequences. Tumour cells may use this mechanism to survive when a medicine blocks a gene required for their growth.

A second gene may take over the original function, allowing the cancer cell to continue operating despite treatment.

The study examines known examples involving breast and prostate cancers. Predicting which genes may act as substitutes could improve understanding of treatment resistance and assist research into drug combinations capable of blocking both pathways.

This potential application also remains at the research stage and does not represent a new treatment currently available to patients.

Could medicines activate compensatory genes?

After identifying and confirming a compensatory gene, the next challenge would be finding a safe way to increase its activity.

Potential future approaches mentioned by the researchers include small molecules and RNA-based technologies. These methods could stimulate a backup gene without directly editing DNA through conventional gene therapy.

Considerable research would still be required. Scientists would need to validate each candidate in relevant cells and tissues, conduct preclinical studies and demonstrate that increasing the gene’s activity does not produce harmful effects.

Limitations of the tool

Genetic compensation may vary according to the tissue, cell type and biological context. A gene capable of replacing another in one type of cell may not perform the same function elsewhere in the body.

The model also depends on the information currently available in biological databases. Extensively researched genes are likely to be better represented than genes about which relatively little is known.

Adding tissue-specific gene-expression information and single-cell data could improve the accuracy of future predictions.

A tool designed to accelerate research

The method’s main contribution is not an immediate diagnosis or treatment. It acts as an intelligent filter to guide experimental research.

The tool could help scientists identify new targets, reduce costs and shorten the initial process of searching for therapeutic mechanisms.

Published in the scientific journal Artificial Intelligence in Medicine, the study positions Malaga researchers at the forefront of an underexplored strategy: using the human genome’s own compensation mechanisms to search for new opportunities against genetic diseases and cancers.