University of Wisconsin–Madison Experts Advance Machine Learning To Boost Cancer Diagnosis And Treatment

Cancer Diagnosis And Treatment: Thanks to machine learning algorithms, short DNA fragments circulating in the bloodstreams of cancer patients can assist physicians in diagnosing specific cancer types and selecting the most effective treatment for a patient.

The new analysis technique, developed by researchers at the University of Wisconsin–Madison and recently published in Annals of Oncology, is compatible with “liquid biopsy” testing apparatus already approved and in use in cancer clinics in the United States. This could expedite the new method’s journey towards assisting patients.

Instead of removing a piece of cancerous tissue from a tumour with a needle, liquid biopsies rely on uncomplicated blood draws.

“Liquid biopsies are much less invasive than tissue biopsies, which, depending on the location of a patient’s tumour, may be impossible in some cases,” says Marina Sharifi, a professor of medicine and oncologist at the UW–Madison School of Medicine and Public Health. Monitoring the status of cancer and the patient’s response to treatment is much simpler if repeated multiple times over the duration of the disease.

University Of Warwick Expert Bags Prestigious National Teaching Award

University of Wisconsin–Madison Experts Advance Machine Learning To Boost Cancer Diagnosis And Treatment

As they develop, cancerous tumours release genetic material called cell-free DNA into the bloodstream. However, not all pieces of a cancer cell’s DNA are liable to break off. A portion of a cell’s DNA is stored in protective spheres called histones. They unwrap sections to access specific portions of the genetic code.

Kyle Helzer, a bioinformatics researcher at the University of Wisconsin–Madison, asserts that portions of the DNA containing the genes utilised frequently by cancer cells are uncoiled more frequently and are therefore more likely to fragment.

Helzer, who is also a co-lead author of the study along with Sharifi and scientist Jamie Sperger, explains, “We’re utilising the greater distribution of these regions in cell-free DNA to identify cancer types.”

The research team, led by UW–Madison senior authors Shuang (George) Zhao, professor of human oncology, and Joshua Lang, professor of medicine, used DNA fragments found in blood samples from a previous study of nearly 200 patients (some with and some without cancer), as well as new samples collected from more than 300 patients treated for breast, lung, prostate or bladder cancers at UW–Madison or other research hospitals in the Big Ten Cancer Research Consortium.

National Surgical Oncologist Day 2023: Date, History, Facts about Cancer

Scientist Analysis on Cancer Diagnosis And Treatment

The scientists divided each sample group into halves. One portion was used to train a machine-learning algorithm to recognise patterns among the fragments of cell-free DNA, which are relatively unique signatures for various types of cancer. The remaining portion was used to verify the trained algorithm. The algorithm translated the results of a liquid biopsy into a cancer diagnosis and the specific varieties of cancer afflicting a patient with an accuracy of over 80 percent.

In addition, the machine learning approach distinguished between two subtypes of prostate cancer: adenocarcinoma, the most common form, and neuroendocrine prostate cancer (NEPC), a fast-progressing variant that is resistant to standard treatment methods. Because NEPC is frequently difficult to distinguish from adenocarcinoma, but necessitates aggressive treatment, it places Lang and Sharifi in a difficult position.

The Research

“Currently, the only way to diagnose NEPC is through a needle biopsy of a tumour site, and even with a high clinical suspicion of NEPC, it can be difficult to obtain a conclusive result using this method,” Sharifi explains.

“Liquid biopsies have advantages in that you do not need to know which tumour site to biopsy, and it is much simpler for the patient to undergo a standard blood draw,” Sperger explains.

Integrated DNA Technologies, based in Iowa, processed the blood samples utilising cell-free DNA sequencing technology. Other methods of “fragmentomic” analysis of cancer DNA in blood samples are supplanted by the use of standard panels, which can reduce the testing time and cost.

“The majority of commercial panels have been designed around the most significant cancer genes that indicate particular drugs for treatment, and they sequence these particular genes,” explains Zhao. We have demonstrated that the same panels and targeted genes can be used to examine the fragmentomics of the cell-free DNA in a blood sample and determine the type of cancer a patient has.

Hundreds of patient samples were collected by the Circulating Biomarker Core and Biospecimen Disease-Oriented Team of the UW Carbone Cancer Centre.