S01.07 The U19 Plans for Integration of Biomarkers Into Future Lung Cancer Screening

Abstract

The goal of the U19 Integrative analysis of Lung Cancer Etiology and Risk (INTEGRAL) consortium is to develop biomarkers that characterize individual risk for development and progression from lung cancer. We are using a comprehensive strategy, depicted below in Figure 1, for this analysis and we are drawing on world-wide resources and expertise. There are three projects focusing on i) genetics of smoking behavior and lung cancer risk, ii) biomarker discovery and validation for identifying individuals at highest risk for developing lung cancer and iii) evaluation of these biomarkers in screening cohorts along with radiographic analaysis to evaluate risk for lung cancer development and nodule behavior. There are also administrative and biostatistics cores. We will discuss strategies and novel findings from these projects. For Project 1, to assist in genetic analysis, we have reimputed all the available data from lung cancer cases and controls using the haplotype reference consortium to bring together a data lake comprising data from over 100,000 individuals. The consortium provides data to its members and to collaborators who would like to evaluate hypotheses related to lung cancer by providing access for analyses and we currently are supporting 107 projects evaluating lung cancer risk. Additionally, consortium members from the University of Laval have performed transcriptomic analysis of normal lung tissue from over 500 participants undergoing surgery for lung cancer treatment. We are also studying the role that genetic factors have in influencing smoking behavior by collaborating with other large consortia and by studying multiethnic variation using Hawaiian multiethnic populations. Analyses of the genetic data and further extension to the UK Biobank have identified novel genetic loci that contribute to risk. Interaction analysis of the CHRNA3/A5/B4 cluster with all other genomic regions identifies interactions with the 15q25.1 nicotinic receptors that influence lung cancer risk. Results identified genes in the neuroactive ligand receptor interaction pathway as playing a key role in increasing lung cancer risk. A cross-ethnicity analysis identified genetic factors in the major histocompatibility complex (MHC) that affect risk for lung cancer. We imputed sequence variation for 26,044 cases and 20,836 controls in classical HLA genes, fine-mapped MHC associations for lung cancer risk with major histologies and compared results among ethnicities. Independent and novel associations within HLA genes were identified in Europeans primarily affecting risk for squamous cell histology including amino acids in the HLA-B*0801 peptide binding groove and an independent HLA-DQB1*06 loci group. In Asians, associations are driven by two independent HLA allele sets affecting adenocarcinoma risk primarily that both increase risk in HLA-DQB1*0401 and HLA-DRB1*0701; the latter was better represented by the amino acid Ala-104. These results implicate several HLA-tumor peptide interactions as the major MHC factor modulating lung cancer susceptibility. A rare variant analysis yielded a mutation of the ATM gene that is rare in all populations except individuals of Jewish descent that primarily increase risk for adenocarcinoma and has highest risk in nonsmoking women. Analyses of smoking and genetic data have identified gene-smoking interactions that contribute to lung cancer risk, and particularly several genes that protect at-risk smokers from lung cancer development. Mendelian randomization and mediation analyses are underway to evaluate novel biomarkers that can be further studied in project 2. This effort found a surprising result that elevated levels of vitamin B12 increase risk for lung cancer development. Project 2 has been bringing together an approach to analyzing biomarkers using data from existing cohort consortia, which have collected samples prior to the clinical presentation of lung cancers. Results of an initial study showed that analysis of 4 circulating proteins (CEA125, CEA, CYFRA 21-1 and pro-SFTB) yielded an area under the receiver operator curve accuracy of 83%. This level of accuracy is sufficient to consider the panel for recruitment of individuals for screening studies, but we anticipate that adding additional biomarkers will further improve the accuracy of risk prediction. Biomarkers that are being further considered include additional protein markers along with micoRNA species, the inclusion of polygenic risk scores and additional serum-derived biomarkers like vitamins B-6 and B-12 that have been shown in mendelian randomization studies to help in identifying high risk subjects. Project 3 is focused on the establishment and validation of the models in the LDCT screening programs. In collaboration with National Lung Screening Trial, Canadian LDCT screening programs, NELSON and United Kingdom Lung Study (UKLS), we have begun the data harmonization across LDCT studies, including clinic-epidemiological data as well as nodule characteristics. We have established a pipeline of feature extractions for the radiomics analysis and compared the inter-reader variability. The intraclass correlation coefficients are >0.75 for the majority of the radiomics features extracted. We will conduct cross-study validation for the model building to ensure the maximum generalizability of the model. We will start the work on biomarkers and assess their added values in these models. biomarkers, early detection, genetics