NEW YORK – New research from a team led by researchers at Peking University People's Hospital suggests that blood lipid profiles can provide clues for detecting early-stage lung cancer.
As they reported in Science Translational Medicine on Wednesday, the investigators linked altered blood lipid levels to early-stage lung cancer, leading to a mass spectrometry-based blood lipid test dubbed the Lung Cancer Artificial Intelligence Detector (LCAID) v2.0.
"This study demonstrates the potential application of LCAID v2.0 for lipidomics-based large-scale population screening, in particularly for populations with high risk of lung cancer," co-senior and co-corresponding authors Yuxin Yin and Jun Wang, researchers at the Peking University Health Science Center and Department of Thoracic Surgery, and their colleagues wrote.
Using 10x Genomics droplet-based single-cell RNA sequencing, the researchers first profiled lipid metabolism-related transcriptional features across nearly 26,700 individual cells isolated from five yet-to-be-treated non-small cell lung cancer tumors, comparing them to new and published scRNA-seq profiles on 55,860 individual cells from eight healthy lung samples.
Along with clusters spanning nine cell lineages in the lung cancer and healthy lung samples, the team identified shifts in lipid metabolism in the NSCLCs. This prompted them to conduct a series of untargeted lipidomic analyses on blood plasma samples from individuals with or without lung cancer.
With high-performance liquid chromatography-mass spectrometry profiling and related support vector machine algorithm analyses on blood samples from 171 individuals with early-stage non-small cell lung cancer and 140 unaffected control individuals, the team flagged nine lipids with apparent ties to early-stage lung cancer — a set they assessed with additional data for 550 more case or control samples.
The LCAID v2.0 set had 100 percent specificity in an initial validation cohort comprised of samples from 99 individuals with cancer and 40 without, the researchers noted.
When they went on to assess the approach using samples collected from 1,036 individuals from a low-dose computed tomography lung cancer screening program at a Beijing hospital and prospectively collected samples from another 109 individuals, it showed 90 percent sensitivity for detecting early-stage lung cancer, with a specificity of 92 percent, the researchers reported, adding that LCAID v2.0 "performs well in the case of early-stage lung cancer detection."
Moreover, the team noted that non-smoking participants with stage I adenocarcinoma were overrepresented in the follow-up cohorts used for the LCAID v2.0 analyses, suggesting that the lipidomic and machine learning method may catch cases that are at a relatively early point in the disease process.
"[T]his work has established a prototypic method of untargeted lipidomics combined with [machine learning] to further refine a disease detection program for targeted lipidomics combined with [machine learning]," the authors wrote.