Virtual Library

Start Your Search

Q. Li

Author of

  • +

    P1.03 - Poster Session 1 - Technology and Novel Development (ID 150)

    • Event: WCLC 2013
    • Type: Poster Session
    • Track: Biology
    • Presentations: 1
    • +

      P1.03-002 - Multiplexed analysis of lung cancer for distinguishing adenocarcinoma from squamous cell carcinoma (ID 2874)

      09:30 - 16:30  |  Author(s): Q. Li

      • Abstract

      Lung cancer is a leading cause of cancer related deaths, 80% of which are classified as non-small cell carcinomas (NSCLC). Differentiating the two main sub-types of NSCLC, adenocarcinoma (AD) from squamous cell carcinoma (SCC) is crucial for therapeutic decision-making. Current methods for characterizing subtypes may involve DAB stains on up to 7 tissue sections, depending on complexity of diagnosis. This process may deplete precious tissue required for molecular studies sequencing and other predictive markers. The goal of the current study was to measure 11 proteins on a single section using a novel multiplexed immunofluorescence (IF) technology (MultiOmyx[TM]) and evaluate performance of analytical workflows in automatic biomarker scoring and in AD, SCC discrimination, with reference to the Pulmotype® test of 5markers

      The protein markers included in the study were comprised of the five antibodies from the Pulmotype® test - Muc1, CK5/6, TRIM29, CEACAM5 and SLC7A5. Six additional markers TTF1, p40, CK7, CK20, p63, NapsinA were selected based on literature reports. These markers were applied to two separate cohorts of NSCLC cases. The entire set of 11 markers was multiplexed on a 378 core tissue microarray (TMA) containing 213 cases of AD or SCC diagnosis (cohort 1). A second 74 core TMA with 50 cases of AD or SCC was stained with the Pulmotype® markers. Manual scores were generated for the immunofluorescence protein images and DAB stained serial tissue sections were used to generate manual ground truth protein expression scores. The first cohort was used to model diagnosis of AD or SCC using an implementation of Breiman and Cutler’s Random Forest and compared to the performance of a previously published lung classifier using manual DAB scores. Image and data analysis algorithms were developed to aid automated biomarker scoring. These algorithms segment the immunofluorescence images into tumor and stromal areas and compute a large number of biomarker-related metrics. Linear regression modeling was used on a down-selected set of metrics to generate automated protein expression scores per biomarker for the second cohort.

      Manual scoring of all 11 targets demonstrated excellent concordance between fluorescence and DAB. Concordance was also demonstrated between manual DAB scores and automated IF metrics for the five Pulmotype® markers with an overall sensitivity and specificity of 95% and 87%, respectively. Statistical modeling indicated that 9 (of the 11) multiplexed markers provided 97% specificity and 90% sensitivity in classifying AD versus SCC. The observed 7% indeterminate rate measures well against the existing published indeterminate rates for the Pulmotype® test (11%) and the classic IHC marker combination TTF-1/p63 (29%).

      Multiplexed analysis of a single tissue section allows maximum use of limited sample and enhanced protein profiling in context of tissue histology. We have shown that differential diagnosis of AD and SCC may be achieved using a multiplexed panel of markers in a single tissue section, when compared to the Pulmotype® test panel. Concordance between fluorescence and DAB shows transferability of the two detection methods. Furthermore, we demonstrated that image and data analysis tools can be applied for consistent automatic biomarker scoring.