The Hospital for Sick Children (SickKids) researchers examined 13,000 distinct cancers to create the first comprehensive comparison of pediatric and adult cancer. They then used a cutting-edge machine-learning algorithm to create a “atlas” of pediatric cancer.
A microscopic examination and the identification of particular proteins are the primary methods used to diagnose cancer in an estimated 18.1 million people annually throughout the world. These techniques’ accuracy varies, and institutes have a difficult time exchanging improvements. This is especially true for pediatric cancer, which is the leading cause of disease-related death in children older than infants in developed countries.
Dr. Adam Shlien, a Senior Scientist in the Genetics & Genome Biology department whose team created this algorithm, says that until new techniques are developed, the complexity of cancer diagnostics will continue to rise as the disease’s prevalence increases globally. Every hospital can utilize the platform to improve the speed and accuracy of cancer diagnosis, even for rare forms.
Study of transcriptome shows uniqueness of pediatric cancer
This machine-learning method classifies every major kind of childhood cancer that is currently understood and can improve, or match, a given cancer diagnosis for 85% of pediatric cancer patients.
This machine-learning technique examines a person’s whole transcriptome as opposed to other methods that may only study the genome, such as cancer panel tests that check for mutations in particular genes. All of a cell’s DNA makes up the genome, but only a subset of it, known as the transcriptome, is copied into RNA molecules.
Dr. Federico Comitani, a Research Associate in the Genetics & Genome Biology program and the study’s first author, explains that just because you have a very active cancer genome doesn’t mean that everything is being acted out in the open. The core characteristics of each tumor can be discovered by comprehensive transcriptome analysis, which also helps to build a more accurate picture of each person’s unique cancer activity.
The vast amount of data gathered by the study team and magnification offered by the technology allowed researchers to identify 455 subtypes of cancer in addition to identifying important distinctions between cancer kinds. The notion that most childhood cancers have a similar origin and later differentiate into a myriad of distinct tumor subtypes is supported by the huge number of subtypes.
For the first time, they could distinguish minor variations between various cancer subtypes. The number of genes expressed in a cell is more variable in childhood tumors than in adult cancers, “Shlien, an Associate Director in the Department of Pediatric Laboratory Medicine and holder of a Canada Research Chair in Childhood Cancer Genomics, explains. This opens up a completely new perspective on cancer, one that may help them determine a cancer’s prognosis and perhaps even alter their perception of the disease.
A classifier can enhance the diagnosis of pediatric cancers
As part of the SickKids Cancer Sequencing program (KiCS), which offers thorough genomic sequencing for children with cancer, the tool is already helping to make cancer diagnoses faster and more accurately.
The categories discovered by this approach predicted significant differences in tumor differentiation and patient survival in cases of neuroblastoma, the most prevalent extracranial solid tumor in children. Similarly, research using the platform revealed an immune cell imbalance that clarified the inconsistent response of sarcomas, cancers of the bone and soft tissue, to immunotherapy and provided information on prospective new therapeutic strategies.
This classifier has the potential to become a standard test for detecting pediatric cancer, adds Shlien, as they continue to add more samples to this expanding atlas and validate it with larger data sets and sample kinds.
A number of early adopter cancer centers across the world are now using this RNA platform for research purposes exclusively. It gives clinicians the ability to obtain a digital diagnosis and compare their patient’s diagnosis to cancer types the platform has discovered. Also, efforts are being made to make this technology available to a larger audience as a platform for diagnostic testing and the quickening of the development of cancer drug products.
