Artificial intelligence enhances the early detection of rare diseases, with its accuracy potentially reaching 90 percent.

Artificial intelligence google deepmind

Even with 90% accuracy, artificial intelligence models can detect genetic changes that may lead to rare diseases. This is crucial because genetic screening and AI analysis could contribute to the detection of these diseases at a pre-symptomatic stage, paving the way for effective treatment when available. It may also enhance the work on identifying factors that can cause various diseases and on innovative therapies. However, for orphan diseases, the scarcity of databases for algorithms to learn from remains an issue.

“Today, artificial intelligence is primarily about analyzing large datasets. The challenge lies in rare diseases whose developmental causes we do not know or cannot detect quickly in typical, preventative scenarios. There lies the greatest potential, provided that the methodology and AI model building are based on exploratory, transdisciplinary assumptions, where we will look for causes of disease development and factors deciding about these diseases’ development. There, the potential for artificial intelligence is the greatest, and there we will also achieve the most progress in understanding and cognitive effect of science in medical areas,” assesses Dr. Eng. Cezary Mazurek, Director of the Poznań Supercomputing and Networking Center, in an interview with the Newseria Innovations agency.

Such tools are already being developed. For example, Google DeepMind recently released a tool that can significantly speed up the process of detecting genetic mutations responsible for rare diseases such as cystic fibrosis or sickle cell anemia. AlphaMissense is a machine learning model that can analyze missense mutations and predict the likelihood of diseases they determine with 90% accuracy. The model operates similarly to language models but has been trained in the language of human biology. It detects improper sequences in the genetic code, just as language models detect improper sentence structure or words that do not fit the meaning.

“The most spectacular, of course, are those tools that can help us understand the mechanism of disease development and early detection of those diseases that are still asymptomatic. Today, this knowledge is based on the analysis of many modalities, many data sources. We analyze the genetic profile, coexisting diseases, various elements related to potential factors of disease development, but we still do not have complete knowledge about it in all diseases,” admits Dr. Eng. Cezary Mazurek. “In those areas where patient data and data obtained at the treatment stage can help doctors ask why this is happening to this patient and not otherwise, that is, to move away from statistical methods and think about the reasons why the disease development does not fit what we know about it, there we should place the most emphasis on the application of artificial intelligence.”

He adds that technologically we are ready to use artificial intelligence methods and IT infrastructure in Poland. We have a very strong background in processing large amounts of data, supercomputing centers, fast fiber-optic networks, and, above all, the competence of programming teams, computer scientists, who are at the forefront of artificial intelligence worldwide.

PCSS participates in creating a unique platform within the MOSAIC project carried out by the Institute of Bioorganic Chemistry, Polish Academy of Sciences. This platform will enable the use of artificial intelligence for conducting innovative research integrating multidimensional biomedical and clinical data to obtain new knowledge and tools for widely available, personalized prevention, diagnostics, and medical therapy.

“The barrier to the development of artificial intelligence is primarily access to anonymized but good quality data. It is also a certain methodology of working with data, working with patients. We need to learn to work on artificial intelligence systems, also in relation to the financing model, building transdisciplinary teams, searching for cross-disciplinary scenarios in other disciplinary areas that we can transfer to medicine,” says the director of the Poznań Supercomputing and Networking Center.

Although more and more is being said about rare diseases, the knowledge base on them is still quite sparse. One of the assumptions of the National Plan for Rare Diseases is to create a register, thanks to which information not only about the number of patients with a given disease entity but also more in-depth data of statistical significance will be the starting point for in-depth work on improving diagnostics and the process of developing new therapies. In mid-October, a new website of the Ministry of Health on rare diseases started. It is intended to be a comprehensive source of information primarily for patients and their families, as well as for health care workers and public institutions.