Neuroblastoma, a deadly childhood cancer, has long defied precise predictions due to its biological complexity.

In a breakthrough study, researchers used a machine learning-driven approach to map out a detailed gene expression landscape of this disease.

Their analysis identified 528 prognostic genes, with 11 emerging as central "hub" markers—including a newly spotlighted gene, RFC3, strongly linked to poor survival outcomes.

Beyond diagnosis, these biomarkers also hinted at responsiveness to existing chemotherapy drugs.

By merging large-scale RNA sequencing with immune profiling and network analysis, the study opens a new path toward more accurate prognostics and tailored treatments for young patients battling neuroblastoma.

Neuroblastoma is the most common solid tumour in infants and accounts for nearly 15% of all paediatric cancer-related deaths.

Despite decades of progress in surgery, chemotherapy, and stem cell therapies, survival for high-risk patients remains under 60%.

Current biomarkers—such as MYCN amplification or ALK mutations—offer limited reach, present only in subsets of patients or requiring complex testing.

These limitations leave a critical gap in effectively predicting disease progression and guiding treatment.

Due to these challenges, there is a pressing need to uncover new, interpretable biomarkers that can improve early risk stratification and drive forward more personalised therapies.

A research team at the Children's Hospital of Chongqing Medical University has unveiled a powerful machine learning framework for identifying prognostic biomarkers in neuroblastoma.

Published in Paediatric Discovery in May 2025, the study leverages bulk and single-cell RNA sequencing data from over 1,200 patients to build a comprehensive prognostic network.

The team's integrative approach not only isolated 11 key gene signatures but also revealed how these genes interact with the tumour microenvironment and drug responses—paving the way for more precise and effective treatment plans in paediatric oncology.

To decipher neuroblastoma's complex genetic architecture, the researchers applied an enhanced version of the stSVM machine learning model to analyse bulk RNA-seq data from 1,207 patients.

This process uncovered 528 genes strongly linked to survival outcomes.

Using weighted gene co-expression network analysis (WGCNA), the team filtered this list to 11 hub genes—AURKA, BLM, BRCA1, BRCA2, CCNA2, CHEK1, E2F1, MAD2L1, PLK1, RAD51, and notably, RFC3.

High expression of RFC3 correlated with poor prognosis and low natural killer (NK) cell activity, hinting at its role in immune evasion.

The study also revealed that tumours with elevated RFC3 expression were more sensitive to vincristine and cyclophosphamide—standard chemotherapy agents.

Further exploration using single-cell RNA sequencing confirmed higher RFC3 expression in epithelial and myeloid cells among short-survival patients, along with reduced T cell infiltration.

These multilayered findings not only highlight RFC3 as a novel biomarker but also suggest it may shape the immune landscape and drug response in neuroblastoma.

By combining gene networks, immune signatures, and drug sensitivity profiles, the research offers a rich, systems-level understanding of the disease.

"Our integrative approach offers a more complete picture of neuroblastoma biology," said Dr. Yupeng Cun, senior investigator of the study.

"Identifying RFC3 as a novel prognostic marker is particularly promising—it not only correlates with patient survival but also with response to key chemotherapies. By merging machine learning with multi-omics data, we've uncovered patterns that traditional analyses often miss. These findings could help clinicians better identify high-risk patients and tailor treatments more effectively, ultimately improving outcomes for children facing this devastating disease."

This study lays critical groundwork for advancing precision medicine in paediatric oncology.

The ability to identify prognostic biomarkers like RFC3—and link them to both immune profiles and drug responsiveness—may transform how neuroblastoma is diagnosed and treated.

In the future, clinicians could use RFC3 expression levels to stratify patients, predict therapeutic response, and guide individualised care.

Furthermore, the study's integrative pipeline could be adapted to other aggressive cancers, making it a valuable tool beyond neuroblastoma.

Continued experimental validation and incorporation of additional omics data will be key to translating these insights into clinical applications that improve survival and quality of life for young patients.

Source: Chongqing Medical University