말초혈액 순환 DNA (circulating cell-free DNA, cfDNA)의 암 시그니처 앙상블(Cancer Signature Ensemble, CSE)을 이용한 CT에서 발견된 폐결절의 암 감별 = Novel cfDNA biomarker for indeterminate pulmonary nodule discrimination : the Cancer Signature Ensemble (CSE)

국문 초록 (Abstract)

검진을 통한 폐암 조기 진단은 금연과 함께 폐암 관련 사망률을 줄이기 위한 가장 유망한 전략이다. 고위험군의 저선량 흉부 컴퓨터단층촬영(LDCT)은 대규모 전향적 연구를 통해서 폐암 생존율 향상의 확실한 근거를 확립하였지만, LDCT의 높은 위양성률은 주요한 제한점이 되고 있다. 본 연구의 목적은 CT에서 발견된 불확정 (Indeterminate) 폐결절의 폐암 여부를 평가하기 위한 혈액 바이오마커(Cancer Signature Ensemble, CSE)를 개발하는 것이다.

본 연구는 다기관, 관찰적, 전향적 검체 수집의 케이스-컨트롤 연구로, CSE는 cfDNA 분석과 후성유전학적 프로파일링의 정보를 인공지능 분석으로 개발하였다. 연구 대상자는 흡연력이 20갑년 이상인 50~80세의 폐암 고위험군 중에서 Lung-RADS 4B/4X 카테고리의 폐 결절이 확인되고 만약 폐암이 진단된 경우 임상 병기 1A기가 예상되는 대상자를 모집하였다. 혈장 cfDNA에서 전체 게놈 메틸화 패턴, DNA 복제 수 변화, DNA 분절화 다양성을 분석하였다. 연구의 1차 목표는 CSE 모델을 개발하고 폐결절 진단력을 평가하는 것이다.

전체 316명의 대상자가 등록되었으며, CSE 모델 개발 코호트에는 최종 184명이 포함되었다(폐암: 116명, 양성 결절: 68명). 메틸화 모델을 개발하기 위해서, 개발 코호트를 다시 트레이닝과 테스트 코호트(6:4)로 나누었다. 트레이닝 코호트에서 기계학습 알고리즘으로 30개의 메틸화 부위를 선별하여 모델을 구축하였다. 그 후 메틸화 모델을 독립된 테스트 코호트에서 검증하였다. DNA 복제수 변이는 cfDNA의 시퀀싱 리드의 개수를 이용하여 기계학습으로 모델을 개발하였다. 최종 CSE 모델은 다양한 조합의 게놈 특징(genomic feature) 모델들을 병합하고 비교한 후 최적의 진단력을 보이는 메틸화 모델과 복제수 변이 모델의 조합으로 선택하였다. 전체 코호트에서 CSE 모델의 폐암 감별 진단력은 ROC (Receiver operating characteristic) 커브의 AUC (Area Under a ROC Curve) 값이 0.85 (95% CI: 0.78-0.91)로 우수한 감별력을 확인하였다. CSE 모델은 대상자의 연령, 결절의 크기, 흡연력 정도, 그리고 결절 침상변연 등의 알려진 폐암 위험 인자들과는 독립적으로 폐암의 위험도를 증명하였다. CSE 모델을 기존의 임상 예측 모델인 Mayo Clinic 모델 또는 Brock 모델과 결합하였을 때(CSE Plus) 단독의 임상 모델들보다 AUC 값이 0.15 상승하는 진단력 향상을 보여주었다(AUC of CSE Plus (Mayo): 0.86; CSE Plus (Brock): 0.87). CSE Plus (Brock) 모델을 특이도 50% (low cutoff)와 특이도 81% (high cutoff)의 기준으로 두 개의 cutoff 값을 정했을 경우 폐결절은 세 가지 위험군으로 분류할 수 있다. 이는 Lung-RADS 지침 대비 약 81% (55/68)의 불필요한 침습적 시술 또는 수술을 줄일 수 있는 효과를 확인하였다.

cfDNA 바이오마커 CSE는 폐암 고위험군 불확정 폐결절 관리 시, 즉각적인 의사결정 과정에서 사용하여, 불필요한 침습적 시술 등의 LDCT 위양성으로 인한 잠재적 위험을 줄이고, 조기 폐암 환자 진단 지연을 예방하는데 유용할 것으로 기대한다. 앞으로, 이를 증명하고 평가하기 위해서 확장된 검증 연구가 필요하다.

다국어 초록 (Multilingual Abstract)

Early diagnosis of lung cancer through screening, along with smoking cessation, is the most promising strategy in the effort to reduce lung cancerrelated mortality. Solid evidence of improved survival rates with screening performed using low-dose chest computed tomography (LDCT) in high-risk groups has been reported from large-scale, prospective studies. However, its high false positive rate is a major limitation of LDCT screening. The objective of our study is to develop a blood biomarker index (Cancer Signature Ensemble,CSE) for estimation of lung cancer probability in screening-detected indeterminate nodules. The CSE was developed through integration of untargeted cfDNA analyses and
epigenetic profiling. Participants aged 50-80 with a smoking history of ≥20 pack years were recruited. Because performance of CSE must be demonstrated in patients with early lung cancer, only patients with lung RADS 4B/4X and expected clinical stage IA if lung cancer was diagnosed were enrolled. Assessment of whole-genome methylation patterns, DNA fragmentation, and copy number variation in cfDNA from plasma was performed. A total of 316 subjects were enrolled, and the final 184 subjects were included in the CSE model development cohort (lung cancer: 116, benign nodules: 68). To develop the methylation model, the development cohort was again divided (6:4) into training and testing cohorts. the methylation model was built by selecting 30 differentially methylated regions from the training cohort using a machinelearning algorithm. The methylation model was then validated in the independent test cohort. For DNA copy number variation, a model was developed using machine learning using the number of cfDNA sequencing reads. The final CSE model was selected as a combination of the methylation model and copy number variation model that showed optimal diagnostic power after merging and comparing various combinations of genomic features. In the entire cohort, the
lung cancer differential diagnostic power of the CSE was confirmed to be excellent, with the AUC (Area Under ROC Curve) value of the ROC (Receiver operating characteristic) curve being 0.85 (95% CI: 0.78-0.91). The CSE model demonstrated the risk of lung cancer independently of known lung cancer risk factors such as the subject's age, nodule size, smoking history, and nodule margin. When the CSE model was combined with the Mayo Clinic model or Brock model, which are existing clinical prediction models, diagnostic power was improved with the AUC value increasing by 0.15 compared to the clinical models alone (AUC of CSE Plus (Mayo): 0.86; CSE Plus (Brock): 0.87). If two cutoff values are set for the CSE Plus (Brock) model based on specificity of 50% (low cutoff) and specificity of 81% (high cutoff), pulmonary nodules can be classified into three risk groups. This can have the effect of reducing unnecessary invasive procedures or surgeries by approximately 81% (55/68) compared to the LungRADS guidelines.We anticipate that CSE will be useful in the process of immediate decisionmaking during management of indeterminate pulmonary nodules, reducing the need for unnecessary invasive procedures or potential risks due to false positives of LDCT, and preventing delayed diagnosis of patients with early lung cancer.
Keywords: lung cancer screening, Biomarker, cfDNA, DNA methylation

목차 (Table of Contents)

• 제1장 서론 1
• 제2장 연구 방법 · 4
• 1. 임상 연구 방법 4
• A. 연구 설계 4
• B. 연구 절차 4
• C. 연구 대상 5
• D. 데이터 수집 6
• 2. 실험실 연구 방법 6
• A. 샘플 수집, 처리 및 준비 6
• B. Whole-genome methylation sequencing 7
• C. NGS 전처리 및 품질 관리 7
• D. 메틸화 분절 정의 · 8
• E. 메틸화 차이 정량화(Average methylation fraction, AMF) · 9
• F. DMR (differentially methylated regions) 검색과 패널 선택 9
• G. CNR (copy number ratio) 모델 개발 · 9
• H. FSR model development 10
• I. CSE model development 10
• J. CSE Plus development 10
• K. Methods for analysis of Model Performance 11
• L. Statistical Analysis Methods 11
• 제3장 결과 13
• 1. 임상 결과 13
• A. 연구 코호트에서 분석 대상자 선택 ·13
• B. 양성 결절과 폐암 대상자 비교 13
• 2. 폐암 관련 임상 인자 ·20
• 3. CSE 모델 개발 과정 ·21
• 4. CSE Plus 모델 개발(훈련 세트) ·26
• 5. 전체 코호트(Combined Set)에서 모델 성능 검증 28
• 6. 하위그룹 분석 ·31
• 7. CSE Plus (brock) 모델의 임상 적용 효과 분석 35
• 제4장 고찰 38
• 1. 연구 결과 정리 38
• 2. CSE 모델 연구의 장점 ·38
• 3. CSE 마커 개발의 중요성 ·40
• 4. 폐암 바이오마커의 특징 정리 40
• 5. 기존 보고된 바이오마커의 성능과 제한점 41
• A. EarlyCDT-Lung 41
• B. MicroRNAs (miRNAs) ·42
• C. cfDNA & Methylation ·43
• 6. cfDNA 후성유전학적 바이오마커 44
• 7. Genome wide 메틸화 분석 ·46
• 8. 연구의 제한점 47
• 제5장 결론 49
• 참고문헌 50
• 영문초록 55

참고문헌 (Reference)

1. Making sense of the ageing methylome, Seale, K. et al., 23(10): p. 585605, , 2022

2. Cigarette smoking and DNA methylation, Lee, K. and, Z. Pausova, Frontiers in Genetics 4, , 2013

3. NCCN guidelines: nonsmall cell lung cancer, National Comprehensive Cancer Network, 2023 Available from https://www. nccn. org/guidelines/guidelinesdetail?category=1&id=1450, , 2023

4. miRTest: a blood test for lung cancer early detection, Montani, F. et al., 107(6): p. djv063, , 2015

5. Biomarkers in Lung Cancer Screening: a Narrative Review, Marmor, H. N. et al., 2023. 5, , 2023

6. Clinical epigenetics: seizing opportunities for translation, M. Esteller, Berdasco, M. and, 20(2): p. 109127, , 2019

7. The Role of Liquid Biopsy in Early Diagnosis of Lung Cancer, Freitas, C. et al., Frontiers in Oncology 2021. 11, , 2021

8. DNA Methylation Cancer Biomarkers: Translation to the Clinic, Locke, W. J. et al., Frontiers in Genetics 10, , 2019

9. Clinical Circulating Tumor DNA Testing for Precision Oncology, Kim, H. and, K. U. Park, Cancer Res Treat 2023. 55(2): p. 351366, , 2023

10. Evaluation of cellfree DNA approaches for multicancer early detection, Jamshidi, A. et al., 40(12): p. 15371549. e12, , 2022

11. Integrating genomic features for noninvasive early lung cancer detection, Chabon, J. J. et al., 580(7802): p. 245251, , 2020

12. The performance of NLST screening criteria in Asian lung cancer patients, Kumar, V. et al., 15(1): p. 916, , 2015

13. 5 Reduced LungCancer Mortality with LowDose Computed Tomographic Screening, The National Lung Screening Trial Research Team, 365(5): p. 395409, , 2011

14. Reduced LungCancer Mortality with Volume CT Screening in a Randomized Trial, de Koning, H. J. et al., 382(6): p. 503513, , 2020

15. Counteracting aged DNA methylation states to combat ageing and agerelated diseases, Reale, A. et al., 206: p. 111695, , 2022

16. Identification of Circulating Tumor DNA for the Early Detection of Smallcell Lung Cancer, FernandezCuesta, L. et al., 10: p. 117123, , 2016

17. Performance of LungRADS in the National Lung Screening Trial: a retrospective assessment, Pinsky, P. F. et al., 162(7): p. 48591, , 2015

18. Lung Screening Benefits and Challenges: A Review of The Data and Outline for Implementation, Sands, J. et al., 16(1): p. 37 53, , 2021

19. Noninvasive early detection of cancer four years before conventional diagnosis using a blood test, Chen, X., et al., 11(1): p. 3475, , 2020

20. Stage Migration and Lung Cancer Incidence After Initiation of LowDose Computed Tomography Screening, Vachani, A. et al., 17(12): p. 13551364, , 2022

21. Liquid biopsies based on DNA methylation as biomarkers for the detection and prognosis of lung cancer, Li, P. et al., 14(1): p. 118, , 2022

22. Earlier diagnosis of lung cancer in a randomised trial of an autoantibody blood test followed by imaging, Sullivan, F. M., et al., 57(1), , 2021

23. Epigenomewide scan identifies differentially methylated regions for lung cancer using prediagnostic peripheral blood, Zhao, N. et al., 17(4): p. 460472, , 2022

24. Global cancer statistics 2020GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries, Sung, H. et al., 71(3): p. 209249, , 2021

25. 28 A Panel of Novel Detection and Prognostic Methylated DNA Markers in Primary NonSmall Cell Lung Cancer and Serum DNA, Ooki, A., et al., 23(22): p. 71417152, , 2017

26. DNA methylation changes measured in prediagnostic peripheral blood samples are associated with smoking and lung cancer risk, Baglietto, L. et al., 140(1): p. 5061, , 2017

27. Prolonged lung cancer screening reduced 10year mortality in the MILD trial: new confirmation of lung cancer screening efficacy, Pastorino, U. et al., 30(7): p. 11621169, , 2019

28. Audit of the autoantibody test, EarlyCDT®Lung, in 1600 patients: An evaluation of its performance in routine clinical practice, Jett, J. R. et al., 83(1): p. 5155, , 2014

29. DNA methylation and associated gene expression in blood prior to lung cancer diagnosis in the Norwegian Women and Cancer cohort, Sandanger, T. M. et al., 8(1): p. 16714, , 2018

30. Mining TCGA and GEO databases for the prediction of poor prognosis in lung adenocarcinoma based on upregulated expression of TNS4, Liu, F. et al., 101(42): p. e31120, , 2022

31. Association of computed tomography screening with lung cancer stage shift and survival in the United States: quasiexperimental study, Potter, A. L., et al., 376: p. e069008, , 2022

32. Baseline computed tomography screening and blood microRNA predict lung cancer risk and define adequate intervals in the BioMILD trial, Pastorino, U. et al., 33(4): p. 395405, , 2022

33. Autoantibody Signature Enhances the Positive Predictive Power of Computed Tomography and NoduleBased Risk Models for Detection of Lung Cancer, Massion, P. P. et al., 12(3): p. 578584, , 2017

34. Clinical utility of a plasmabased miRNA signature classifier within computed tomography lung cancer screening: a correlative MILD trial study, Sozzi, G., et al., 32(8): p. 76873, , 2014

35. DNA Methylation in Lung Cancer: Mechanisms and Associations with Histological Subtypes, Molecular Alterations, and Major Epidemiological Factors, Hoang, P. H. and, M. T. Landi, 14(4): p. 961, , 2022

36. Validation of Lung EpiCheck, a novel methylationbased blood assay, for the detection of lung cancer in European and Chinese highrisk individuals, Mina, G., et al., 57(1): p. 2002682, , 2021

37. Can autoantibody tests enhance lung cancer screening?—an evaluation of EarlyCDT ® Lung in context of the German Lung Cancer Screening Intervention Trial (LUSI), González Maldonado, S. et al., 10(1): p. 233242, , 2021

38. Accurate classification of pulmonary nodules by a combined model of clinical, imaging, and cellfree DNA methylation biomarkers: a model development and external validation study, He, J., et al., 2023, , 2023

39. Assessment of Plasma Proteomics Biomarker’s Ability to Distinguish Benign From Malignant Lung Nodules: Results of the PANOPTIC (Pulmonary Nodule Plasma Proteomic Classifier) Trial, Silvestri, G. A. et al., 154(3): p. 491500, , 2018