• Find people
• Intranet

News

12/07/2024 - Press release

UPF and Hospital del Mar design a method to classify different types of knee osteoarthritis with greater precision using advanced technologies

The new method would allow for a more objective and precise classification of patients' types of osteoarthritis, supported by data mining techniques and biological modeling, which enable the analysis and processing of large volumes of information.

Pompeu Fabra University (UPF) and Hospital del Mar have designed a new method to classify the types of knee osteoarthritis that patients suffer from in a more objective and precise way. This method is supported by data mining and machine learning techniques coupled with modeling of the biological regulation of cells. These techniques make it possible to cross-reference real data from patients' joints with augmented data obtained from computational biology model simulations.

This research is part of the STRATO research project funded by the Spanish State Research Agency, led by Prof. Jérôme Noailly and Dr. Simone Tassani (UPF), and by Drs. Joan Carlos Monllau and Jordi Monfort from Hospital del Mar and researchers from its research institute. It is also part of the O-Health project, led by Jérôme Noailly and funded by the European Research Council (ERC). This particular study was co-directed by Noailly and Dr. Gemma Piella (UPF).

Currently, the main method for diagnosing knee osteoarthritis primarily relies on a combination of imaging techniques and symptoms expressed by patients-conditioned by subjective perceptions-with the support of medical questionnaires. Furthermore, in most cases, the disease is detected in advanced stages when it is only possible to alleviate its consequences but not reverse its effects.

The new method will detect dysfunctions in cartilage tissue production that cause knee osteoarthritis, contributing to a more objective understanding of the symptoms that patients suffer, which is also highly relevant in precision medicine.

The study, recently published in an article in the journal Scientific Reports (Nature), provides significant advances for one of the most prevalent diseases in contemporary societies. Osteoarthritis affects 22% of men and 31% of women over 55, according to the Institute for Health Metrics and Evaluation (2019), and its incidence is expected to increase in the future.

Computational simulations to detect whether the cartilage-producing cells in the joints, or chondrocytes, function normally

The new method has been designed and tested on a sample of 51 female patients with moderate knee osteoarthritis. On the one hand, it continues to use clinical data that define the pathological state or symptoms of each patient. On the other hand, it allows for the correlation of real clinical data from each patient with simulated data on the behavior of their chondrocytes: the cells responsible for producing cartilage, the cushioning tissue that covers joints and helps reduce friction between the connecting bones. If a patient's chondrocytes exhibit destructive behavior, osteoarthritis will be triggered, and if they exhibit normal or preservative behavior, the disease will not develop.

The behavior of each patient's chondrocytes can be calculated from a personalized computational model that relates the proteins in the knee's synovial fluid to those found inside the chondrocytes. It is a network-based computational model that integrates a vast amount of knowledge about the biology of articular cartilage to simulate mechanobiological behaviors. The network nodes represent proteins and other molecules important for regulating chondrocyte activity and their interactions.

With this computational model, the specific behavior of each patient's chondrocytes can be simulated based on the proteins detected in the synovial fluid. Therefore, it has great potential in the field of prevention, as it would allow for the detection of the causes of osteoarthritis from its earliest stages.

Once both the clinical data and the simulated data of each patient's knee are available, they can be correlated thanks to the third element of the diagnostic method: the use of a data analytics and machine learning model called SVM (Support Vector Machine). As Jérôme Noailly, head of the Biomechanics and Mechanobiology (BMMB) research area at BCN MedTech at UPF, explains: "We perform a data mining exercise with machine learning, in our case with SVM technology. It allows us to evaluate which molecules and mechanisms indicated by the model and simulations are best related to the pathology and its clinical descriptors in each patient."

Towards the identification of new osteoarthritis biomarkers

The correlation between the clinical and biological data of people suffering from osteoarthritis can also contribute to the identification of biomarkers for this pathology. Biomarkers are essentially substances that indicate the presence of biological elements or processes that serve various purposes, including diagnosing a disease, defining its severity and extent, or evaluating the effectiveness of a treatment. In the case of osteoarthritis, there is currently no validated biomarker recognized by the scientific community that serves as a therapeutic target, meaning to determine which specific part of the body should receive pharmacological treatment for the disease.

To advance in the detection of such biomarkers, the research followed this procedure. First, patients were divided into two different groups based on whether they presented higher or lower levels for seven descriptors of osteoarthritis symptoms. Then, it was examined which potential biomarker (or distinguishing element) had the most influence in classifying the patients into groups.

One of the potential biomarkers identified is the level of proteins in the knee's synovial fluid. "While we have also examined potential biomarkers present in urine or blood, we consider that the proteomics of synovial fluid may be more effective for diagnosing osteoarthritis," explains Maria Segarra-Queralt, the first author of the research article and a researcher at BCN MedTech at UPF. It should be noted that synovial fluid is found within the joints and, therefore, has a much closer proximity to cartilage than substances found in urine or blood. In the future, these and other potential biomarkers identified by the study will need to be validated with a larger number of patients.

Among the most notable results, which are still preliminary, the researchers have been able to confirm that current clinical data alone do not explain the inflammation measured in the joint. The hypothesis is that this is due to current clinical data being influenced by each patient's pain perception, which, in turn, depends on each individual's sociocultural situation. Therefore, it is proposed to limit the use of current clinical data, which do not provide objective information about the disease's origin, and instead measure the protein levels in the knee fluid.

Thanks to this study, it has been found that the body's innate defenses have a significant influence on the development of osteoarthritis. This could have a positive impact on the future treatment of the disease. For example, if a patient experiences a lot of pain but has low levels of these proteins, it would indicate that their pain is altered by other pathways not directly stemming from osteoarthritis, such as depression that alters an individual's pain perception. Therefore, it would likely be counterproductive for this patient to undergo surgery to replace their knee with a prosthesis. Finally, thanks to the use of the computational model, the behavior of chondrocytes has been determined. The results indicate that promoting the biological regeneration of chondrocytes would have greater therapeutic potential than treatments focused on suppressing the pro-inflammatory signals of these cells.

Reference article

Segarra-Queralt, M., Galofré, M., Tio, L. et al. Characterization of clinical data for patient stratification in moderate osteoarthritis with support vector machines, regulatory network models, and verification against osteoarthritis Initiative data. Sci Rep 14, 11797 (2024). https://doi.org/10.1038/s41598-024-62212-x

More news