Artificial intelligence is changing the way researchers investigate diseases. It can analyse large volumes of biological information, identify patterns that would be difficult to detect manually, and help prioritise potential drug candidates for further study.
In rare disease research, these capabilities are particularly valuable. Researchers often work with limited patient populations, making every experiment and every dataset especially important.
However, there is one essential point that is sometimes overlooked.
Artificial intelligence cannot generate meaningful discoveries without reliable biological data.
Before AI can identify potential therapeutic opportunities, scientists first need high-quality laboratory models that accurately represent the diseases they want to study. They also need carefully generated experimental data that capture the biological mechanisms underlying those diseases.
This is why biology remains the foundation of AI-driven drug discovery.
AI depends on high-quality biology
Artificial intelligence is exceptionally good at recognising patterns in complex datasets.
It can compare thousands of molecular interactions, analyse gene expression profiles, identify relationships between proteins and predict which compounds deserve further investigation.
What AI cannot do is generate biological evidence on its own.
It cannot determine whether a disease model faithfully reproduces the biology observed in patients. It cannot assess whether an experiment was performed consistently. And it cannot compensate for incomplete, inconsistent or poor-quality datasets.
The quality of AI predictions depends directly on the quality of the biological information used to train and validate the algorithms.
Simply put, better biological data leads to better scientific insights.
Why rare diseases present a unique challenge
More than 7,000 rare diseases have been identified worldwide, yet only a small proportion currently have approved treatments. One of the reasons is that research is often limited by the availability of biological material.
For many rare diseases, only a small number of patients are available for study. This means researchers have fewer samples, fewer opportunities to generate experimental data and fewer biological models than they would for more common diseases.
The DREAMS project addresses this challenge by focusing initially on five rare neuromuscular disorders that share common biological characteristics related to autophagy dysfunction and desmin disorganisation. These shared features provide researchers with an opportunity to investigate common disease mechanisms rather than studying each condition in isolation.
Building reliable disease models
One of the key technologies used in DREAMS is induced pluripotent stem cells, commonly known as iPSCs.
These cells can be generated from patient-derived samples and then differentiated into skeletal muscle cells. Because they retain the patient’s genetic background, they provide researchers with laboratory models that closely reflect important aspects of disease biology.
According to the DREAMS Description of Action, these cellular models will support several objectives, including identifying shared biomarkers, investigating disease mechanisms and performing high-throughput drug screening across the selected neuromuscular disorders.
Producing these models requires carefully standardised laboratory procedures to ensure that experiments are reproducible and that the resulting datasets are suitable for computational analysis.
Only once this biological foundation has been established can AI begin analysing the data.
From laboratory experiments to biological datasets
Modern biomedical research generates many different types of biological information.
For example, researchers may analyse:
- Gene expression profiles
- Protein abundance
- Cellular responses to potential medicines
- Microscopy images
- Functional measurements of cell behaviour
Each dataset provides a different perspective on how a disease affects cells.
Individually, these datasets are informative.
Combined, they provide a much more comprehensive understanding of disease biology.
Within DREAMS, transcriptomic and proteomic analyses are combined with phenotypic drug screening to generate complementary biological datasets. These data will support AI-based analyses aimed at identifying shared drug targets and prioritising potential therapeutic strategies across multiple rare neuromuscular disorders.
AI helps researchers identify meaningful patterns
Once high-quality biological datasets have been generated, AI becomes a powerful analytical tool.
Rather than replacing experimental research, AI helps scientists interpret complex biological information more efficiently.
For example, computational models may help researchers:
- Identify biological pathways shared by multiple diseases.
- Prioritise promising drug targets.
- Suggest existing medicines that could potentially be investigated for additional indications.
- Highlight other diseases that may share similar biological mechanisms.
Importantly, these computational predictions do not represent confirmed discoveries.
Within DREAMS, AI-generated hypotheses are followed by experimental validation using cellular models and later preclinical studies. This stepwise workflow ensures that computational predictions are evaluated through biological evidence before any future clinical development is considered.
Why reproducibility matters
Producing large amounts of biological data is not enough.
Researchers also need data that are reproducible.
If the same experiment is performed under the same conditions, it should produce comparable results regardless of when or where it is carried out.
This consistency is essential because machine learning algorithms identify statistical patterns. If datasets contain excessive experimental variation, AI models may learn laboratory-specific noise instead of genuine biological relationships.
For this reason, standardised experimental protocols remain a cornerstone of biomedical research.
AI is one part of a much larger scientific process
Artificial intelligence sometimes receives attention as though it independently discovers new medicines.
In reality, successful biomedical research depends on expertise from many different scientific disciplines.
Cell biologists develop laboratory models.
Clinicians contribute medical knowledge and patient expertise.
Data scientists develop computational methods.
Bioinformaticians integrate complex biological datasets.
Drug discovery specialists evaluate promising therapeutic candidates.
Each discipline contributes essential knowledge that supports the next stage of research.
AI strengthens this process by helping researchers extract meaningful information from carefully generated biological datasets.
Looking ahead
The combination of advanced laboratory models and artificial intelligence has the potential to improve how researchers investigate rare diseases.
The DREAMS project explores how high-quality biological datasets generated from patient-derived cellular models can support AI analyses that identify shared disease mechanisms and prioritise future therapeutic strategies. These activities remain part of the project’s ongoing research programme, and any candidate compounds identified will require further experimental validation before potential clinical application.
As biomedical technologies continue to evolve, one principle is unlikely to change.
Artificial intelligence can accelerate discovery, but meaningful scientific progress still begins with high-quality biological research.
