Genetic map produced in Brazil points to new avenues for treating Alzheimer’s

By Roseli Andrion  |  FAPESP Innovative R&D – For decades, the search for an Alzheimer’s disease treatment has focused on a few molecular targets, such as the tau and beta-amyloid proteins that accumulate in the brains of people with the disease. However, this list is beginning to expand with the creation of genetic maps of the disease, such as one developed by BioDecision Analytics.

Using artificial intelligence and transcriptomics, a field of study that analyzes the genetic instructions used by cells, researchers at the startup in São Paulo, Brazil, have identified 129 new potential therapeutic targets. This breakthrough opens up new avenues of research for drug development and diagnosis of the disease via blood test. Alzheimer’s disease affects more than 57 million people worldwide, including nearly 2 million in Brazil, and there is still no cure or effective treatment.

“One of the only known targets for developing Alzheimer’s disease therapies was beta-amyloid [a protein that accumulates abnormally in the brain and forms plaques that block communication between neurons and trigger inflammatory processes]. That’s why it was imperative to explore other avenues,” Rodrigo Araldi, one of the founders of BioDecision, tells FAPESP Innovative R&D. “Focusing on a single target helps explain why so many drugs have failed in recent decades,” the researcher notes.

To carry out the mapping, the researchers used 2,870 brain tissue and blood samples from healthy individuals and patients diagnosed with Alzheimer’s disease, all of whom had passed away. The samples were sourced from the Sequence Read Archive (SRA), a public-domain database maintained by the U.S. government.

The brain samples covered five distinct regions – the hippocampus, the cingulate cortex, Wernicke’s area, the visual cortex, and Broca’s area – which are vital for memory, language, vision, and hearing functions. According to Araldi, segmentation is crucial because the brain is not homogeneous. “Areas linked to memory, such as the hippocampus, have genetic signatures that are completely distinct from those of areas responsible for vision or hearing,” he explains. “Mixing these regions into a single analysis would be a methodological error that would obscure essential details.”

Technology and precision

The data processing utilized the BDASeq® transcriptomic analysis platform, which was developed by the startup with support from FAPESP’s Innovative Research in Small Businesses Program (PIPE). This tool analyzes RNA (ribonucleic acid) sequencing to identify differential gene expression, i.e., which genes are “turned on” or “turned off” at a given moment. Since RNA guides cells in producing proteins, mapping it allows researchers to determine whether a gene is being activated more or less intensely in the context of the disease.

“Most studies use a single statistical method to do this, which generates a lot of noise,” notes João Rafael Dias Pinto, BioDecision’s co-founder and co-lead on the analysis. “Our platform combines eight established techniques to filter out what is truly relevant. That eliminates false positives, which lead researchers down the wrong paths, and false negatives, which cause valuable clues to be discarded.”

In addition to statistical rigor, the platform features an artificial intelligence layer that cross-references genetic data with clinical information. “We don’t just list altered genes. With the help of AI, we select those with the greatest biological relevance and concrete potential to become drugs or diagnostic tools,” the researcher explains.

What the map reveals

The analysis identified more than 4,200 genes with altered expression in Alzheimer’s disease patients compared to healthy individuals in the same age group. Taking age and sex into account is one of the pillars of the study. “Comparing the genetic profile of an 80-year-old patient with that of a 20-year-old is like comparing the engine of a used car with that of a brand-new one,” Dias Pinto explains. “Ignoring these variables leads to conclusions that don’t hold up in real life.”

The analysis allowed for the classification of two main groups among the findings. The first group consists of 75 genes associated with processes such as neuronal plasticity, or the ability of neurons to form new connections (synapses), which is lost in Alzheimer’s. These genes serve as promising therapeutic targets and include eight that were previously unreported in the scientific literature. The second group consists of 74 genes that are altered in both the brain and blood. These genes are ideal candidates for biomarkers, or biological markers of disease progression. Of these, 21 genes demonstrated over 90% accuracy in distinguishing between affected and healthy individuals. This suggests that the presence of these alterations is correlated with Alzheimer’s.

“Today, the diagnosis of Alzheimer’s is clinical, time-consuming, and imprecise,” says Araldi. “Tests such as PET [positron emission tomography] scans cost tens of thousands of reais and aren’t available in the public healthcare system. With this discovery, we’ve opened the door to blood tests for diagnosis, which are much more accessible.” In this case, the presence of the disease would be inferred by the presence of RNA molecules in the blood associated with altered gene expression.

The study was presented in March at the AD/PD Congress (International Conference on Alzheimer’s and Parkinson’s Diseases), the world’s leading scientific event focused on neurodegenerative diseases, held in Copenhagen, Denmark.

Proof of concept and democratization

Before studying Alzheimer’s, BioDecision validated BDASeq® using Huntington’s disease. The study received awards at the MENA Congress on Rare Diseases in Abu Dhabi, United Arab Emirates, for three consecutive years (2024, 2025, and 2026). The biomarkers were so significant that a Brazilian pharmaceutical company acquired the data and is now developing a stem cell therapy (read more at pesquisaparainovacao.fapesp.br/3135).

Starting in August 2026, BioDecision will make the platform widely accessible through a paid, cloud-based subscription supported by Google Cloud. “We want smaller laboratories to be able to perform cutting-edge analyses without needing their own supercomputers,” says Araldi. In addition to FAPESP, the development of the technology received support from the Brazilian Federation of Alzheimer’s Associations (FEBRAZ) and Alzheimer’s Disease International (ADI).

Araldi’s message to the pharmaceutical industry is pragmatic: “We’ve done the hardest part – mining the data. Now, the industry needs to show genuine interest in identifying new targets. We need to stop insisting on what has failed so that we can finally offer dignity and quality time to millions of families.”

* Additional images and information about research supported by FAPESP can be found on the Foundation’s Flickr profile