For the first time, researchers have identified at least three unique subtypes of a rare bone cancer, potentially transforming clinical trials and patient care.
A research project led by the University of East Anglia (UEA) has used advanced mathematical modeling and a machine learning technique called “Latent Process Decomposition” (LPD) to categorize patients with osteosarcoma into different subgroups based on their genetic data. Previously, all patients were grouped together and treated using the same protocols, resulting in very mixed outcomes.
While genetic sequencing has helped uncover subtypes of other cancers, such as breast or skin cancer—enabling targeted, personalized treatments—this has been much harder to achieve for osteosarcoma. This rare cancer starts in the bone and typically affects children and teenagers.
Lead author Dr. Darrell Green of UEA’s Norwich Medical School explained, “Since the 1970s, osteosarcoma has been treated using untargeted chemotherapy and surgery, which can lead to limb amputation and severe, lifelong side effects from chemotherapy.
“Over the last 50-plus years, multiple international clinical trials investigating new drugs for osteosarcoma have been deemed ‘failures.’ However, our research shows that in each of these ‘failed’ trials, a small subset of patients—around five to 10 percent—responded to the new drugs, suggesting the presence of osteosarcoma subtypes that could benefit from these treatments.
“The new medicines were not total ‘failures’ as previously concluded. Instead, the drugs were unsuccessful for every patient with osteosarcoma but could have been effective for select subgroups.
“We hope that grouping patients using this new algorithm will lead to successful clinical trial outcomes for the first time in over half a century. When patients are treated with targeted drugs specific to their cancer subtype, we can move away from standard chemotherapy.”
The search for kinder, more targeted treatments for osteosarcoma is a priority for Children with Cancer UK, a leading childhood cancer charity. In 2021, the charity awarded funding to the UEA team to explore innovative treatment approaches for osteosarcoma.
Dr. Sultana Choudhry, Head of Research at Children with Cancer UK, stated, “Investing in pioneering research programs is integral to achieving our vision of a world where every child and young person survives cancer.
“By funding groundbreaking research, we are not only advancing scientific knowledge but also finding gentler, more effective treatments for our youngest and most vulnerable cancer patients. We hope the outcomes of this research will improve the diagnosis, treatment, and long-term care of young cancer patients.”
The survival rate for osteosarcoma has stagnated at around 50% for the past 45 years. This is due to limited understanding of the cancer’s subtypes, how the immune system interacts with the tumor, and the factors contributing to treatment resistance or cancer spread.
Scientists have yet to identify the key biological markers that predict patient outcomes or treatment responses. These gaps in knowledge hinder progress in improving survival rates.
Previous efforts to predict osteosarcoma subtypes using computational methods suggested the presence of distinct subtypes. However, these models assumed that each tumor could be neatly placed into one group, overlooking the variation within individual tumors, which often comprise diverse cancer cell types.
In this study, researchers used the more advanced LPD method, which considers the heterogeneity within tumors. Unlike earlier approaches, LPD identifies hidden patterns in gene activity, representing different “functional states” of the tumor, each with its specific gene expression profile.
The LPD method determines how many of these patterns are needed to describe a particular tumor. The study uncovered three osteosarcoma subtypes, one of which responded poorly to the standard chemotherapy regimen known as MAP. By grouping patients based on these patterns, doctors can make more informed treatment decisions.
Researchers acknowledged limitations, including the small dataset used to develop the LPD model and incomplete clinical data in the validation cohort. Access to tissue samples and linked clinical data is particularly challenging for osteosarcoma due to its rarity, limited biopsy material, and chemotherapy-related damage in post-treatment samples.
Despite these challenges, the LPD method consistently identified subgroups across four independent datasets, demonstrating its reliability. Like any machine learning tool, its accuracy will improve as more data becomes available.
(ANI)
