118P - Whole transcriptome sequencing of lung tissue to combine disease classification and identification of actionable targets

Background

In the era of precision oncology, the histopathological cancer classification is complemented by additional molecular classification. We sought to evaluate the capability of whole transcriptome sequencing (WTS) of lung tumors and lung diseases to (i) reproduce the histopathological classification and (ii) identify actionable targets.

Methods

We performed WTS of 294 formalin-fixed paraffin-embedded (FFPE) tissue samples resected from the lung including 5 primary cancer types, 13 metastasis types, as well as sarcoidosis, tuberculosis, and healthy lung. One-against-all classifiers were fitted by gene filtering followed by logistic regression and integrated in a multiclass classifier. The classifier predicted one of 19 classes or delivered “no prediction” in case of inconclusive prediction. The classifier was validated using cross-validation and in an independent data set of 321 tumors of 15 cancer types from the TCGA. Differentially expressed genes (DEGs) were identified by the Wilcoxon test followed by correction of p-values using the Benjamini-Hochberg method. Sets of DEGs were functionally analyzed by investigating the enrichment of the cancer hallmark categories from MSigDB.

Results

Cross-validation resulted in 87.5% correctly classified, 11.3% unclassified (“no prediction”), and 1.2% misclassified samples. The independent validation resulted in 78.5% correctly classified, 19.0% unclassified, and 2.5% misclassified samples. For many of the unclassified samples, the result could be explained by the inconclusive separation of lung squamous cell carcinomas (SCC) and metastasis of other SCC. Analysis of DEGs and gene set enrichment analysis revealed differences between primary tumors, metastases, and non-malignant diseases. Gene expression profiles of tyrosine kinase inhibitors (TKIs) and antibody-drug conjugates (ADCs) were investigated for targetable alterations.

Conclusions

WTS of clinical lung tissue samples was feasible and a gene expression-based multiclass predictor reproduced the histopathological tissue classification with high accuracy. In future, evaluation of the mRNA expression of genes targetable by TKIs and ADCs could contribute to the guidance of targeted therapies.

Editorial acknowledgement

Clinical trial identification

Legal entity responsible for the study

The authors.

Funding

Qlucore AB, Lund, Sweden.

Disclosure

C. Brorsson, C. Ivarrson, J. Ripa, F. Hellborg: Financial Interests, Personal, Full or part-time Employment: Qlucore. J. Rade: Financial Interests, Personal, Project Lead: Qlucore. A. Stenzinger: Financial Interests, Personal, Advisory Role: Qlucore. All other authors have declared no conflicts of interest.