Lung cancer remains one of the leading causes of cancer death worldwide, and survival is closely tied to the stage at diagnosis. Although low-dose computed tomography (LDCT) screening has been proven to significantly reduce lung cancer mortality, its implementation in real-world settings still faces many challenges. For example, among people eligible for screening in the United States, actual LDCT participation remains below 10%, while many European countries have yet to establish large-scale systematic screening programs. Preserving diagnostic performance while improving the accessibility and adherence of screening has become an urgent challenge in the field of lung cancer early screening.

A joint research team from Germany and the United States has published important findings in the Journal of Thoracic Oncology, proposing miLung — an early lung cancer detection method based on whole-blood small RNA sequencing. By building a diagnostic model from small RNA expression features in blood, the study offers a convenient, non-invasive and scalable new strategy for lung cancer screening.

For a long time, the non-coding regions of the genome were dismissed as “junk DNA.” Yet research in recent years has continually shown that these regions in fact carry rich regulatory information; the various non-coding RNAs in particular play important roles in the onset and progression of disease.

In this study, the research team conducted a prospective study in 1,384 high-risk individuals who met NLST screening criteria (aged 55–74 years, with a smoking history of ≥30 pack-years). Using ultra-deep small RNA sequencing (approximately 20 million sequencing reads per sample), the researchers systematically profiled small RNA expression in whole blood and ultimately built a lung cancer diagnostic model composed of 18 small RNA features — miLung.

The study used PAXgene whole blood RNA tubes, which lyse whole blood immediately after collection and stabilize both intracellular and extracellular RNA. No cap opening, pipetting or centrifugation is needed at the clinical site, fundamentally avoiding the pre-analytical batch variability commonly seen in conventional liquid biopsy.

The team developed a specific antisense oligonucleotide blocking method targeting 20 high-abundance miRNAs and 35 other small RNAs, with a median blocking efficiency of 91%. This breakthrough freed sequencing bandwidth from a state of being “flooded,” increasing the number of detected miRNA species by  28% (p=0.0014) and bringing previously masked low-abundance signals to the surface.

The study used full-resolution isomiR and small RNA variant expression data, rather than limiting itself to canonical miRNAs, thereby retaining richer biological information.

Development of the whole-blood small RNA sequencing workflow

Built on this technical framework, the miLung model developed by the research team showed stable detection performance across multiple cohorts.

In the 943-participant discovery cohort (445 lung cancer patients and 498 controls), the model achieved a median AUC of 0.86 (95% CI: 0.84–0.86). In the 441-participant independent validation cohort (286 lung cancer patients and 155 controls), the AUC remained at 0.83, indicating good generalizability.

Notably, detection performance improved with advancing tumor stage: the AUC was approximately 0.73–0.76 for stage I lung cancer, while reaching 0.86–0.90 in patients with stage IV disease.

The model was equally stable across different pathological subtypes, with particularly strong performance in lung squamous cell carcinoma (AUC 0.88) and small cell lung cancer (AUC 0.93).

AUC values in the discovery and validation cohorts

Surprisingly, the most heavily weighted feature in the model was a 28S rRNA fragment (28S-rRNA-bin-162), located in theribosomal L1 stalk domain — a region that is essential for tRNA translocation on the ribosome.

Localization of 28S-rRNA-bin-162 and the L1-stalk structure

Conventional ctDNA testing has a sensitivity of only 20%–50% in early-stage lung cancer, mainly because early tumors release extremely small amounts of DNA into the blood. The innovation of miLung lies in simultaneously capturing small RNAs derived from both tumor cells and immune cells, integrating tumor signals with the host immune response to produce an amplifying effect of “1+1>2.”

As the accompanying commentary “Small RNAs Do It Better” pointed out: gathering information simultaneously from the tumor, the microenvironment and the whole organism — especially the immune system — may well be the more promising direction for the future of liquid biopsy.

Compared with imaging-based screening, the miLung test offers clear advantages in accessibility. It can be completed with a routine blood draw during a standard health check-up, with no need to schedule a separate imaging examination. The team also further explored dried blood spot sampling, opening up the possibility of at-home collection and remote screening in the future.

If further validated in larger population studies, this kind of blood-based test could — much as Cologuard changed the model for colorectal cancer screening — help reduce lung cancer mortality at the population level by substantially raising screening participation.