Artificial intelligence and care

Applied AI in research: how synthetic data supports rare diseases and protects privacy but clear rules are needed

The data artificially created by an algorithm mimic the main characteristics of the real starting data but do not belong to real persons or entities, thus allowing the possibility of conducting R&S even on neglected pathologies and bypassing the issue of confidentiality, but a regulatory framework is still awaited to make them fully usable

by Paolo Gasparini *

26 January 2026

3' min read

Translated by AI

Versione italiana

3' min read

Translated by AI

Versione italiana

66% of Italian citizens declare themselves willing to have their health data analysed by artificial intelligence systems to contribute to medical research, while 60% say they are in favour of using AI in scientific research and clinical trials. These are the data that emerged from a survey conducted by Youtrend as part of the Net Health - Sanità in Rete 2030 project, which paints an interesting picture of the relationship between users and technological innovation in healthcare. There is also a cross-party consensus on the political level: all the parliamentarians interviewed and 90% of the regional councillors declared themselves in favour of the use of AI in pharmaceutical research.

"Frontier" summary data

These data suggest that there is fertile ground for the adoption of innovative tools that can accelerate clinical research, improve the efficiency of the healthcare system and protect patient privacy at the same time. Among these tools, synthetic data represent one of the most promising frontiers. The recent participation in the Working Group on Synthetic Data, set up within the Net Health project, promoted and coordinated by LS Cube, provided an opportunity for an in-depth discussion with experts from the scientific, regulatory, legal and economic fields, confirming the relevance and topicality of the topic.

L’identikit

But what exactly is synthetic data? They are data artificially created by an algorithm, which mimic the main characteristics of the real source data, but do not belong to real persons or entities.

What are the major advantages of synthetic data? The solution of privacy problems, the possibility of data augmentation where there is a small number of patients (see rare diseases), the correction of possible bias in 'real world' data. The potential spin-offs make it possible to accelerate medical research, support drug development, and in a broader sense promote personalised medicine.

The applications

One of the areas where synthetic data can develop their full potential is that of rare diseases, diseases that affect a limited number of people (1 in every 1000/2000) and present specific problems in relation to their rarity. Rare diseases are almost all of genetic origin and lack an effective therapeutic perspective. In this context, summary data can contribute to:

Accelerare drug development. For example, in the case of lysosomal storage disease (Gaucher), synthetic data showed a high degree of realism, being suitable for predicting patients' response to treatment.

Promuovere and simplify the conduct of clinical trials by simulating the placebo effect with 'synthetic patients' and thus reducing the need to include real patients in control groups (often impractical due to numerical shortage and ethical aspects).

Favorire therapeutic customisation, through increasingly accurate identification of individual variability in response to treatments.

Enhanced privacy

Strengthen the protection of patient privacy. The generation of synthetic datasets reduces the likelihood that sensitive information can be traced back to specific individuals.

Data augmentation. Starting with only 19 patients suffering from an ultra-rare neurodevelopmental disorder (White-Sutton syndrome), the synthetic data generated not only presented a close overlap with the original phenotypic data but also improved the ability to predict the severity of cognitive level (IQ).

What prospects

Are we ready to use synthetic data? While great strides have been made from a scientific/technological perspective, the application of synthetic data in everyday medical practice is currently still limited by the need to a) define clear standards for their generation and validation, b) ensure their quality and statistical accuracy, c) develop a suitable regulatory framework and corresponding governance.

* Representative for Italy of the Committee for Medicinal Products for Human Use (Chmp) of the Ema; Professor of Medical Genetics at the University of Trieste; Director of Advanced Diagnostic Services and Medical Genetics at the Irccs Materno Infantile "Burlo Garofolo" of Trieste