Rayleen Bowman

OBJECTIVE: To classify automatically lung tumor-node-metastases (TNM) cancer stages from free-text pathology reports using symbolic rule-based classification. DESIGN: By exploiting report substructure and the symbolic manipulation of systematized nomenclature of medicine-clinical terms (SNOMED CT) concepts in reports, statements in free text can be evaluated for relevance against factors relating to the staging guidelines. Post-coordinated SNOMED CT expressions based on templates were defined and populated by concepts in reports, and tested for subsumption by staging factors. The subsumption results were used to build logic according to the staging guidelines to calculate the TNM stage. MEASUREMENTS: The accuracy measure and confusion matrices were used to evaluate the TNM stages classified by the symbolic rule-based system. The system was evaluated against a database of multidisciplinary team staging decisions and a machine learning-based text classification system using support vector machines. RESULTS: Overall accuracy on a corpus of pathology reports for 718 lung cancer patients against a database of pathological TNM staging decisions were 72%, 78%, and 94% for T, N, and M staging, respectively. The system's performance was also comparable to support vector machine classification approaches. CONCLUSION: A system to classify lung TNM stages from free-text pathology reports was developed, and it was verified that the symbolic rule-based approach using SNOMED CT can be used for the extraction of key lung cancer characteristics from free-text reports. Future work will investigate the applicability of using the proposed methodology for extracting other cancer characteristics and types.

Disease progression of chronic obstructive pulmonary disease (COPD) is variable, with some patients having a relatively stable course, while others suffer relentless progression leading to severe breathlessness, frequent acute exacerbations of COPD (AECOPD), respiratory failure and death. Radiological markers such as CT emphysema index, bronchiectasis and coronary artery calcification (CAC) have been linked with increased mortality in COPD patients. Molecular changes in lung tissue reflect alterations in lung pathology that occur with disease progression; however, lung tissue is not routinely accessible. Cell counts (including neutrophils) and mediators in induced sputum have been associated with lung function and risk of exacerbations. Examples of peripheral blood biological markers (biomarkers) include those associated with lung function (reduced CC-16), emphysema severity (increased adiponectin, reduced sRAGE), exacerbations and mortality [increased CRP, fibrinogen, leukocyte count, IL-6, IL-8, and tumor necrosis factor α (TNF-α)] including increased YKL-40 with mortality. Emerging approaches to discovering markers of gene-environment interaction include exhaled breath analysis [volatile organic compounds (VOCs), exhaled breath condensate], cellular and systemic responses to exposure to air pollution, alterations in the lung microbiome, and biomarkers of lung ageing such as telomere length shortening and reduced levels of sirtuins. Overcoming methodological challenges in sampling and quality control will enable more robust yet easily accessible biomarkers to be developed and qualified, in order to optimise personalised medicine in patients with COPD.

PURPOSE: Improving outcomes for early-stage lung cancer is a major research focus at present because a significant proportion of stage I patients develop recurrent disease within 5 years of curative-intent lung resection. Within tumor stage groups, conventional prognostic indicators currently fail to predict relapse accurately. EXPERIMENTAL DESIGN: To identify a gene signature predictive of recurrence in primary lung adenocarcinoma, we analyzed gene expression profiles in a training set of 48 node-negative tumors (stage I-II), comparing tumors from cases who remained disease-free for a minimum of 36 months with those from cases whose disease recurred within 18 months of complete resection. RESULTS: Cox proportional hazards modeling with leave-one-out cross-validation identified a 54-gene signature capable of predicting risk of recurrence in two independent validation cohorts of 55 adenocarcinomas [log-rank P=0.039; hazard ratio (HR), 2.2; 95% confidence interval (95% CI), 1.1-4.7] and 40 adenocarcinomas (log-rank P=0.044; HR, 3.3; 95% CI, 1.4-7.9). Kaplan-Meier log-rank analysis found that predicted poor-outcome groups had significantly shorter survival, and furthermore, the signature predicted outcome independently of conventional indicators of tumor stage and node stage. In a subset of earliest stage adenocarcinomas, generally expected to have good outcome, the signature predicted samples with significantly poorer survival. CONCLUSIONS: We describe a 54-gene signature that predicts the risk of recurrent disease independently of tumor stage and which therefore has potential to refine clinical prognosis for patients undergoing resection for primary adenocarcinoma of the lung.

MicroRNAs (miRNAs) are a family of small, non-coding RNA species functioning as negative regulators of multiple target genes including tumour suppressor genes and oncogenes. Many miRNA gene loci are located within cancer-associated genomic regions. To identify potential new amplified oncogenic and/or deleted tumour suppressing miRNAs in lung cancer, we inferred miRNA gene dosage from high dimensional arrayCGH data. From miRBase v9.0 (http://microrna.sanger.ac.uk), 474 human miRNA genes were physically mapped to regions of chromosomal loss or gain identified from a high-resolution genome-wide arrayCGH study of 132 primary non-small cell lung cancers (NSCLCs) (a training set of 60 squamous cell carcinomas and 72 adenocarcinomas). MiRNAs were selected as candidates if their immediately flanking probes or host gene were deleted or amplified in at least 25% of primary tumours using both Analysis of Copy Errors algorithm and fold change (≥ ± 1.2) analyses. Using these criteria, 97 miRNAs mapped to regions of aberrant copy number. Analysis of three independent published lung cancer arrayCGH datasets confirmed that 22 of these miRNA loci showed directionally concordant copy number variation. MiR-218, encoded on 4p15.31 and 5q35.1 within two host genes (SLIT2 and SLIT3), in a region of copy number loss, was selected as a priority candidate for follow-up as it is reported as underexpressed in lung cancer. We confirmed decreased expression of mature miR-218 and its host genes by qRT-PCR in 39 NSCLCs relative to normal lung tissue. This downregulation of miR-218 was found to be associated with a history of cigarette smoking, but not human papilloma virus. Thus, we show for the first time that putative lung cancer-associated miRNAs can be identified from genome-wide arrayCGH datasets using a bioinformatics mapping approach, and report that miR-218 is a strong candidate tumour suppressing miRNA potentially involved in lung cancer.