The clinical, molecular, and prognostic features of the 2022 WHO and ICC classification systems for myelodysplastic neoplasms.

Myelodysplastic neoplasms (MDS) are clonal disorders of bone marrow failure exhibiting a variable risk of progression to acute myeloid leukemia. MDS exhibit certain prognostic genetic or cytogenetic abnormalities, an observation that has led to both the pathologic reclassification of MDS in the 2022 World Health Organization (WHO) and International Consensus Classification (ICC) systems, as well as to an updated prognostic schema, the Molecular International Prognostic Scoring System (IPSS-M). This single-institution study characterized the molecular patterns and clinical outcomes associated with the 2022 WHO and ICC classification schemas to assess their clinical utility. Strikingly, with the exception of one individual, all 210 patients in our cohort were classified into analogous categories by the two pathologic/diagnostic schemas. Most patients (70%) were classified morphologically while the remaining 30% had genetically classified disease by both criteria. Prognostic risk, as assessed by the IPSS-M score was highest in patients with MDS with biallelic/multi-hit TP53 mutations and lowest in pts with MDS-SF3B1. Median leukemia-free survival (LFS) was shortest for those with MDS with biallelic/multi-hit TP53 (0.7 years) and longest for those with MDS with low blasts (LFS not reached). These data demonstrate the clear ability of the 2022 WHO and ICC classifications to organize MDS patients into distinct prognostic risk groups and further show that both classification systems share more similarities than differences. Incorporation of the IPSS-M and IPSS-R features provide additive prognostic and survival components to both the WHO and ICC classifications, which together enhance their utility for evaluating and treating MDS patients.

Overall Survival Among Patients With De Novo Stage IV Metastatic and Distant Metastatic Recurrent Non-Small Cell Lung Cancer.

Importance: Despite recent breakthroughs in therapy, advanced lung cancer still poses a therapeutic challenge. The survival profile of patients with metastatic lung cancer remains poorly understood by metastatic disease type (ie, de novo stage IV vs distant recurrence). Objective: To evaluate the association of metastatic disease type on overall survival (OS) among patients with non-small cell lung cancer (NSCLC) and to identify potential mechanisms underlying any survival difference. Design, Setting, and Participants: Cohort study of a national US population based at a tertiary referral center in the San Francisco Bay Area using participant data from the National Lung Screening Trial (NLST) who were enrolled between 2002 and 2004 and followed up for up to 7 years as the primary cohort and patient data from Stanford Healthcare (SHC) for diagnoses between 2009 and 2019 and followed up for up to 13 years as the validation cohort. Participants from NLST with de novo metastatic or distant recurrent NSCLC diagnoses were included. Data were analyzed from January 2021 to March 2023. Exposures: De novo stage IV vs distant recurrent metastatic disease. Main Outcomes and Measures: OS after diagnosis of metastatic disease. Results: The NLST and SHC cohort consisted of 660 and 180 participants, respectively (411 men [62.3%] vs 109 men [60.6%], 602 White participants [91.2%] vs 111 White participants [61.7%], and mean [SD] age of 66.8 [5.5] vs 71.4 [7.9] years at metastasis, respectively). Patients with distant recurrence showed significantly better OS than patients with de novo metastasis (adjusted hazard ratio [aHR], 0.72; 95% CI, 0.60-0.87; P < .001) in NLST, which was replicated in SHC (aHR, 0.64; 95% CI, 0.43-0.96; P = .03). In SHC, patients with de novo metastasis more frequently progressed to the bone (63 patients with de novo metastasis [52.5%] vs 19 patients with distant recurrence [31.7%]) or pleura (40 patients with de novo metastasis [33.3%] vs 8 patients with distant recurrence [13.3%]) than patients with distant recurrence and were primarily detected through symptoms (102 patients [85.0%]) as compared with posttreatment surveillance (47 patients [78.3%]) in the latter. The main finding remained consistent after further adjusting for metastasis sites and detection methods. Conclusions and Relevance: In this cohort study, patients with distant recurrent NSCLC had significantly better OS than those with de novo disease, and the latter group was associated with characteristics that may affect overall survival. This finding can help inform future clinical trial designs to ensure a balance for baseline patient characteristics.

Identification and confirmation via in situ hybridization of Merkel cell polyomavirus in rare cases of posttransplant cutaneous T-cell lymphoma.

BACKGROUND: Viral infection is an oncogenic factor in many hematolymphoid malignancies. We sought to determine the diagnostic yield of aligning off-target reads incidentally obtained during targeted hematolymphoid next-generation sequencing to a large database of viral genomes to screen for viral sequences within tumor specimens. METHODS: Alignment of off-target reads to viral genomes was performed using magicBLAST. Localization of Merkel cell polyomavirus (MCPyV) RNA was confirmed by RNAScope in situ hybridization. Integration analysis was performed using Virus-Clip. RESULTS: Four cases of post-cardiac-transplant folliculotropic mycosis fungoides (fMF) and one case of peripheral T-cell lymphoma (PTCL) were positive in off-target reads for MCPyV DNA. Two of the four cases of posttransplant fMF and the case of PTCL showed localization of MCPyV RNA to malignant lymphocytes, whereas the remaining two cases of posttransplant fMF showed MCPyV RNA in keratinocytes. CONCLUSIONS: Our findings raise the question of whether MCPyV may play a role in rare cases of T-lymphoproliferative disorders, particularly in the skin and in the heavily immunosuppressed posttransplant setting.

Genomic landscape of T-large granular lymphocyte leukemia and chronic lymphoproliferative disorder of NK cells: a single institution experience.

LGLL is a rare and chronic lymphoproliferative disorder including T-LGLL and CLPD-NK. Here, we investigated the genomic profiles of LGLL with a focus on STAT3 and STAT5B mutations in a cohort of 49 patients (41 T-LGLL, 8 CLPD-NK). Our study indicated that STAT3 was identified in 38.8% (19/49) of all patients, while STAT5B occurred in only 8.2% (4/49) of patients. We found that STAT3 mutations were associated with lower ANC in T-LGLL patients. The average number of pathogenic/likely pathogenic mutations in STAT3/STAT5B-mutated patients was significantly higher than that in WT patients (1.78 ± 1.17 vs 0.65 ± 1.36, p = 0.0032). Additionally, TET2-only mutated T-LGLL (n = 5) had a significant reduction in platelet values compared with the WT (n = 16) or STAT3-only mutated T-LGLL (n = 12) (p < 0.05). In conclusion, we compared the somatic mutational landscape between STAT3/STAT5B WT and mutated patients and correlate with their distinct clinical characteristics.

Minimal/Measurable Residual Disease Monitoring in Patients with Lymphoid Neoplasms by High-Throughput Sequencing of the T-Cell Receptor.

High-throughput sequencing of the T-cell receptor beta (TRB) and gamma (TRG) loci is increasingly utilized due to its high sensitivity, specificity, and versatility in the diagnosis of various T-cell malignancies. Application of these technologies for tracking disease burden can be valuable in detecting recurrence, determining response to therapy, guiding future management of patients, and establishing endpoints for clinical trials. In this study, the performance of the commercially available LymphoTrack high-throughput sequencing assay was assessed for determining residual disease burden in patients with various T-cell malignancies. A custom bioinformatics pipeline and database was also developed to facilitate minimal/measurable residual disease analysis and clinical reporting. This assay demonstrated excellent test performance characteristics, achieving a sensitivity of 1 of 100,000 T-cell equivalents for the DNA inputs evaluated and high concordance with orthogonal testing methods. This assay was further utilized to correlate disease burden in several patients, demonstrating its potential utility for monitoring patients with T-cell malignancies.

Tracing Founder Mutations in Circulating and Tissue-Resident Follicular Lymphoma Precursors.

Follicular lymphomas (FL) are characterized by BCL2 translocations, often detectable in blood years before FL diagnosis, but also observed in aging healthy individuals, suggesting additional lesions are required for lymphomagenesis. We directly characterized early cooperating mutations by ultradeep sequencing of prediagnostic blood and tissue specimens from 48 subjects who ultimately developed FL. Strikingly, CREBBP lysine acetyltransferase (KAT) domain mutations were the most commonly observed precursor lesions, and largely distinguished patients developing FL (14/48, 29%) from healthy adults with or without detected BCL2 rearrangements (0/13, P = 0.03 and 0/20, P = 0.007, respectively). CREBBP variants were detectable a median of 5.8 years before FL diagnosis, were clonally selected in FL tumors, and appeared restricted to the committed B-cell lineage. These results suggest that mutations affecting the CREBBP KAT domain are common lesions in FL cancer precursor cells (CPC), with the potential for discriminating subjects at risk of developing FL or monitoring residual disease. SIGNIFICANCE: Our study provides direct evidence for recurrent genetic aberrations preceding FL diagnosis, revealing the combination of BCL2 translocation with CREBBP KAT domain mutations as characteristic committed lesions of FL CPCs. Such prediagnostic mutations are detectable years before clinical diagnosis and may help discriminate individuals at risk for lymphoma development. This article is highlighted in the In This Issue feature, p. 1275.

Predictive Model to Guide Brain Magnetic Resonance Imaging Surveillance in Patients With Metastatic Lung Cancer: Impact on Real-World Outcomes.

PURPOSE: Brain metastasis is common in lung cancer, and treatment of brain metastasis can lead to significant morbidity. Although early detection of brain metastasis may improve outcomes, there are no prediction models to identify high-risk patients for brain magnetic resonance imaging (MRI) surveillance. Our goal is to develop a machine learning-based clinicogenomic prediction model to estimate patient-level brain metastasis risk. METHODS: A penalized regression competing risk model was developed using 330 patients diagnosed with lung cancer between January 2014 and June 2019 and followed through June 2021 at Stanford HealthCare. The main outcome was time from the diagnosis of distant metastatic disease to the development of brain metastasis, death, or censoring. RESULTS: Among the 330 patients, 84 (25%) developed brain metastasis over 627 person-years, with a 1-year cumulative brain metastasis incidence of 10.2% (95% CI, 6.8 to 13.6). Features selected for model inclusion were histology, cancer stage, age at diagnosis, primary site, and RB1 and ALK alterations. The prediction model yielded high discrimination (area under the curve 0.75). When the cohort was stratified by risk using a 1-year risk threshold of > 14.2% (85th percentile), the high-risk group had increased 1-year cumulative incidence of brain metastasis versus the low-risk group (30.8% v 6.1%, P < .01). Of 48 high-risk patients, 24 developed brain metastasis, and of these, 12 patients had brain metastasis detected more than 7 months after last brain MRI. Patients who missed this 7-month window had larger brain metastases (58% v 33% largest diameter > 10 mm; odds ratio, 2.80, CI, 0.51 to 13) versus those who had MRIs more frequently. CONCLUSION: The proposed model can identify high-risk patients, who may benefit from more intensive brain MRI surveillance to reduce morbidity of subsequent treatment through early detection.

Trichodysplasia Spinulosa Polyomavirus Endothelial Infection, California, USA.

We describe 3 patients in California, USA, with trichodysplasia spinulosa polyomavirus (TSPyV) infection of endothelium after steroid administration. We detected TSPyV RNA in tissue specimens by in situ hybridization, which revealed localization to endothelial cells. These cases suggest that diseases associated with endothelial inflammation could be associated with TSPyV infection.

Genomic Profiling of Bronchoalveolar Lavage Fluid in Lung Cancer.

Genomic profiling of bronchoalveolar lavage (BAL) samples may be useful for tumor profiling and diagnosis in the clinic. Here, we compared tumor-derived mutations detected in BAL samples from subjects with non-small cell lung cancer (NSCLC) to those detected in matched plasma samples. Cancer Personalized Profiling by Deep Sequencing (CAPP-Seq) was used to genotype DNA purified from BAL, plasma, and tumor samples from patients with NSCLC. The characteristics of cell-free DNA (cfDNA) isolated from BAL fluid were first characterized to optimize the technical approach. Somatic mutations identified in tumor were then compared with those identified in BAL and plasma, and the potential of BAL cfDNA analysis to distinguish lung cancer patients from risk-matched controls was explored. In total, 200 biofluid and tumor samples from 38 cases and 21 controls undergoing BAL for lung cancer evaluation were profiled. More tumor variants were identified in BAL cfDNA than plasma cfDNA in all stages (P < 0.001) and in stage I to II disease only. Four of 21 controls harbored low levels of cancer-associated driver mutations in BAL cfDNA [mean variant allele frequency (VAF) = 0.5%], suggesting the presence of somatic mutations in nonmalignant airway cells. Finally, using a Random Forest model with leave-one-out cross-validation, an exploratory BAL genomic classifier identified lung cancer with 69% sensitivity and 100% specificity in this cohort and detected more cancers than BAL cytology. Detecting tumor-derived mutations by targeted sequencing of BAL cfDNA is technically feasible and appears to be more sensitive than plasma profiling. Further studies are required to define optimal diagnostic applications and clinical utility. SIGNIFICANCE: Hybrid-capture, targeted deep sequencing of lung cancer mutational burden in cell-free BAL fluid identifies more tumor-derived mutations with increased allele frequencies compared with plasma cell-free DNA. See related commentary by Rolfo et al., p. 2826.

Diagnostic Impact of Next-Generation Sequencing Panels for Lymphoproliferative Neoplasms on Small-Volume Biopsies.

OBJECTIVES: We investigated the feasibility and utility of next-generation sequencing (NGS)-based targeted somatic mutation panels and IG/TR gene rearrangement assays in the diagnosis of lymphoproliferative disorders (LPDs) in small-volume biopsies. MATERIALS: We performed a retrospective, single-institution review of all NGS assays requested over a 3-year period by hematopathologists for diagnostic purposes on small-volume biopsies. RESULTS: We identified 59 small-volume biopsies. The TR assay was most commonly requested (42 [71%]), followed by the somatic mutation panel (32 [54%]) and IG assay (26 [44%]). NGS studies were associated with a change in the diagnostic line in about half of cases (28 [47%]) and in a change in the likelihood of a diagnosis in a further 16 cases (27%); there was no diagnostic impact of NGS testing in 15 cases (25%). CONCLUSIONS: Implementation of NGS panel somatic mutation or IG/TR gene rearrangement assays on small-volume biopsies contributes to the diagnosis of LPDs in the majority of select cases for diagnostic purposes. The molecular diagnosis is considered in the context of the clinical, histologic, and immunophenotypic findings and does not by itself lead to a definitive diagnosis in small-volume biopsies.

Inferring gene expression from cell-free DNA fragmentation profiles.

Profiling of circulating tumor DNA (ctDNA) in the bloodstream shows promise for noninvasive cancer detection. Chromatin fragmentation features have previously been explored to infer gene expression profiles from cell-free DNA (cfDNA), but current fragmentomic methods require high concentrations of tumor-derived DNA and provide limited resolution. Here we describe promoter fragmentation entropy as an epigenomic cfDNA feature that predicts RNA expression levels at individual genes. We developed 'epigenetic expression inference from cell-free DNA-sequencing' (EPIC-seq), a method that uses targeted sequencing of promoters of genes of interest. Profiling 329 blood samples from 201 patients with cancer and 87 healthy adults, we demonstrate classification of subtypes of lung carcinoma and diffuse large B cell lymphoma. Applying EPIC-seq to serial blood samples from patients treated with PD-(L)1 immune-checkpoint inhibitors, we show that gene expression profiles inferred by EPIC-seq are correlated with clinical response. Our results indicate that EPIC-seq could enable noninvasive, high-throughput tissue-of-origin characterization with diagnostic, prognostic and therapeutic potential.

Accurate Detection and Quantification of FLT3 Internal Tandem Duplications in Clinical Hybrid Capture Next-Generation Sequencing Data.

FLT3 internal tandem duplications (ITDs) are found in approximately one-third of patients with acute myeloid leukemia and have important prognostic and therapeutic implications that have supported their assessment in routine clinical practice. Conventional methods for assessing FLT3-ITD status and allele burden have been primarily limited to PCR fragment size analysis because of the inherent difficulty in detecting large ITD variants by next-generation sequencing (NGS). In this study, we assess the performance of publicly available bioinformatic tools for the detection and quantification of FLT3-ITDs in clinical hybridization-capture NGS data. We found that FLT3_ITD_ext had the highest overall accuracy for detecting FLT3-ITDs and was able to accurately quantify allele burden. Although all other tools evaluated were able to detect FLT3-ITDs reasonably well, allele burden was consistently underestimated. We were able to significantly improve quantification of FLT3-ITD allelic burden independent of the detection method by utilizing soft-clipped reads and/or ITD junctional sequences. In addition, we show that identifying mutant reads by previously identified junctional sequences further improves the sensitivity of detecting FLT3-ITDs in post-treatment samples. Our results demonstrate that FLT3-ITDs can be reliably detected in clinical NGS data using available bioinformatic tools. We further describe how accurate quantification of FLT3-ITD allele burden can be added on to existing clinical NGS pipelines for routine assessment of FLT3-ITD status in patients with acute myeloid leukemia.

Performance Characteristics of Mutational Signature Analysis in Targeted Panel Sequencing.

CONTEXT.­: Mutational signatures have been described in the literature and a few centers have implemented pipelines for clinical reporting. OBJECTIVE.­: To describe the performance of a mutational signature caller with clinical samples sequenced on a targeted next-generation sequencing panel with a small genomic footprint. DESIGN.­: One thousand six hundred eighty-two clinical samples were analyzed for the presence of mutational signatures using deconstructSigs on variant calls with at least 20 variant reads. RESULTS.­: Signature 10 (associated with POLe mutation) achieved separation of cases and controls in hypermutated samples. Signatures 4 (associated with tobacco smoking) and 7 (associated with ultraviolet radiation) as an indicator of pulmonary or cutaneous primary sites showed moderate sensitivity and high specificity at optimal cutpoints. Mutational signatures in malignancies with unknown primaries were somewhat consistent with the clinically suspected primary site, with an apparent dose-response relationship between the number of variants analyzed and the ability of mutational signature analysis to correctly suggest a primary site. CONCLUSIONS.­: Mutational signatures represent an opportunity for orthogonal testing of primary site, which may be particularly useful in supporting cutaneous or pulmonary sites in poorly differentiated neoplasms. Tobacco smoking, ultraviolet radiation, and POLe mutational signatures are the most appropriate signatures for implementation. Even relatively small numbers of variants appear capable of supporting a clinically suspected primary.

Machine Learning Predictability of Clinical Next Generation Sequencing for Hematologic Malignancies to Guide High-Value Precision Medicine.

Advancing diagnostic testing capabilities such as clinical next generation sequencing methods offer the potential to diagnose, risk stratify, and guide specialized treatment, but must be balanced against the escalating costs of healthcare to identify patient cases most likely to benefit from them. Heme-STAMP (Stanford Actionable Mutation Panel for Hematopoietic and Lymphoid Malignancies) is one such next generation sequencing test. Our objective is to assess how well Heme-STAMP pathological variants can be predicted given electronic health records data available at the time of test ordering. The model demonstrated AUROC 0.74 (95% CI: [0.72, 0.76]) with 99% negative predictive value at 6% specificity. A benchmark for comparison is the prevalence of positive results in the dataset at 58.7%. Identifying patients with very low or very high predicted probabilities of finding actionable mutations (positive result) could guide more precise high-value selection of patient cases to test.

Noninvasive Early Identification of Therapeutic Benefit from Immune Checkpoint Inhibition.

Although treatment of non-small cell lung cancer (NSCLC) with immune checkpoint inhibitors (ICIs) can produce remarkably durable responses, most patients develop early disease progression. Furthermore, initial response assessment by conventional imaging is often unable to identify which patients will achieve durable clinical benefit (DCB). Here, we demonstrate that pre-treatment circulating tumor DNA (ctDNA) and peripheral CD8 T cell levels are independently associated with DCB. We further show that ctDNA dynamics after a single infusion can aid in identification of patients who will achieve DCB. Integrating these determinants, we developed and validated an entirely noninvasive multiparameter assay (DIREct-On, Durable Immunotherapy Response Estimation by immune profiling and ctDNA-On-treatment) that robustly predicts which patients will achieve DCB with higher accuracy than any individual feature. Taken together, these results demonstrate that integrated ctDNA and circulating immune cell profiling can provide accurate, noninvasive, and early forecasting of ultimate outcomes for NSCLC patients receiving ICIs.

KEAP1/NFE2L2 Mutations Predict Lung Cancer Radiation Resistance That Can Be Targeted by Glutaminase Inhibition.

Tumor genotyping is not routinely performed in localized non-small cell lung cancer (NSCLC) due to lack of associations of mutations with outcome. Here, we analyze 232 consecutive patients with localized NSCLC and demonstrate that KEAP1 and NFE2L2 mutations are predictive of high rates of local recurrence (LR) after radiotherapy but not surgery. Half of LRs occurred in tumors with KEAP1/NFE2L2 mutations, indicating that they are major molecular drivers of clinical radioresistance. Next, we functionally evaluate KEAP1/NFE2L2 mutations in our radiotherapy cohort and demonstrate that only pathogenic mutations are associated with radioresistance. Furthermore, expression of NFE2L2 target genes does not predict LR, underscoring the utility of tumor genotyping. Finally, we show that glutaminase inhibition preferentially radiosensitizes KEAP1-mutant cells via depletion of glutathione and increased radiation-induced DNA damage. Our findings suggest that genotyping for KEAP1/NFE2L2 mutations could facilitate treatment personalization and provide a potential strategy for overcoming radioresistance conferred by these mutations. SIGNIFICANCE: This study shows that mutations in KEAP1 and NFE2L2 predict for LR after radiotherapy but not surgery in patients with NSCLC. Approximately half of all LRs are associated with these mutations and glutaminase inhibition may allow personalized radiosensitization of KEAP1/NFE2L2-mutant tumors.This article is highlighted in the In This Issue feature, p. 1775.

Validated limited gene predictor for cervical cancer lymph node metastases.

PURPOSE: Recognizing the prognostic significance of lymph node (LN) involvement for cervical cancer, we aimed to identify genes that are differentially expressed in LN+ versus LN- cervical cancer and to potentially create a validated predictive gene signature for LN involvement. MATERIALS AND METHODS: Primary tumor biopsies were collected from 74 cervical cancer patients. RNA was extracted and RNA sequencing was performed. The samples were divided by institution into a training set (n = 57) and a testing set (n = 17). Differentially expressed genes were identified among the training cohort and used to train a Random Forest classifier. RESULTS: 22 genes showed > 1.5 fold difference in expression between the LN+ and LN- groups. Using forward selection 5 genes were identified and, based on the clinical knowledge of these genes and testing of the different combinations, a 2-gene Random Forest model of BIRC3 and CD300LG was developed. The classification accuracy of lymph node (LN) status on the test set was 88.2%, with an Area under the Receiver Operating Characteristic curve (ROC-AUC) of 98.6%. CONCLUSIONS: We identified a 2 gene Random Forest model of BIRC3 and CD300LG that predicted lymph node involvement in a validation cohort. This validated model, following testing in additional cohorts, could be used to create a reverse transcription-quantitative polymerase chain reaction (RT-qPCR) tool that would be useful for helping to identify patients with LN involvement in resource-limited settings.

Is Merkel Cell Carcinoma of Lymph Node Actually Metastatic Cutaneous Merkel Cell Carcinoma?

OBJECTIVES: The possibility of a so-called primary lymph node neuroendocrine carcinoma has been described in the literature. Here we evaluate cases fitting such a diagnosis and find that the cases demonstrate a convincing and pervasive pattern consistent with metastatic Merkel cell carcinoma. METHODS: Six cases of primary lymph node Merkel cell carcinoma and one case of metastatic neuroendocrine carcinoma at a bony site, all with unknown primary, were sequenced using a combination of whole-exome and targeted panel methods. Sequencing results were analyzed for the presence of an ultraviolet (UV) mutational signature or off-target detection of Merkel cell polyomavirus (MCPyV). RESULTS: Four of six primary lymph node cases were positive for a UV mutational signature, with the remaining two cases positive for off-target alignment of MCPyV. One case of neuroendocrine carcinoma occurring at a bony site was also positive for a UV mutational signature. CONCLUSIONS: We find no evidence to corroborate the existence of so-called primary Merkel cell carcinoma of lymph node.

Integrating genomic features for non-invasive early lung cancer detection.

Radiologic screening of high-risk adults reduces lung-cancer-related mortality1,2; however, a small minority of eligible individuals undergo such screening in the United States3,4. The availability of blood-based tests could increase screening uptake. Here we introduce improvements to cancer personalized profiling by deep sequencing (CAPP-Seq)5, a method for the analysis of circulating tumour DNA (ctDNA), to better facilitate screening applications. We show that, although levels are very low in early-stage lung cancers, ctDNA is present prior to treatment in most patients and its presence is strongly prognostic. We also find that the majority of somatic mutations in the cell-free DNA (cfDNA) of patients with lung cancer and of risk-matched controls reflect clonal haematopoiesis and are non-recurrent. Compared with tumour-derived mutations, clonal haematopoiesis mutations occur on longer cfDNA fragments and lack mutational signatures that are associated with tobacco smoking. Integrating these findings with other molecular features, we develop and prospectively validate a machine-learning method termed 'lung cancer likelihood in plasma' (Lung-CLiP), which can robustly discriminate early-stage lung cancer patients from risk-matched controls. This approach achieves performance similar to that of tumour-informed ctDNA detection and enables tuning of assay specificity in order to facilitate distinct clinical applications. Our findings establish the potential of cfDNA for lung cancer screening and highlight the importance of risk-matching cases and controls in cfDNA-based screening studies.