Abstract 2774: Reference free transcriptomic characterization of chemoresistant triple negative breast cancers provides a promising reservoir of predictive biomarkers of early chemoresistance

Abstract

Abstract Triple Negative Breast Cancers (TNBCs) are the most aggressive breast cancers and so far, benefited very little from personalized medicine. Unlike in hormone positive breast cancers, coding transcriptomic signatures like Lehmann’s failed to predict which TNBC patients will benefit from perioperative chemotherapy. Here, we analyzed the total transcriptome of a unique cohort of 44 TNBC tumors before neoadjuvant chemotherapy (NAC), including both patients that were chemosensitive (N=26) and chemoresistant (N=18) to following treatment. Differential gene expression analysis (DE-seq) was first performed on annotated genes and de novo Scallop RNA-profiled transcripts. Nearly half of detected differential transcripts were from non-coding genes that were mostly long non-coding (lnc)RNAs (90%). More than half differential lncRNAs were not annotated in gencode. Next, De-kupl, a new reference-free analysis of differential fragments of transcripts, was applied. This approach is based on 31-nucleotide long k-mers that are later merged into contigs without annotation bias. These differential contigs were able to separate correctly, in an unsupervised manner, all TNBC patients regarding their response to chemotherapy, outperforming our previous DE-seq analysis of gencode and scallop transcripts. Strikingly, 80% of the DE-kupl differential contigs were not derived from DE-seq differential genes. Furthermore, as BRCA 1/2 mutated TNBCs are now well known for their chemosensitivity, we excluded them from subsequent analysis and repeated the differential expression analysis. Our results showed that DE-kupl differential contigs once more perfectly classified all patients and outperformed the whole length transcripts from DE-seq. Finally, we applied the LASSO regression analysis machine learning method, on both all misregulated genes and all misregulated contigs in parallel, to generate minimal signatures characterizing pre-chemotherapy TNBC tumors that are going to be chemoresistant. We obtained an 8-contig-signature that overpassed the 11-gene-signature in the accuracy to discriminate chemosensitive and chemoresistant TNBC patients before chemotherapy. These results suggest that a contig based unreferenced differential analysis of TNBCs regarding their response status to NAC provides promising predictive biomarkers of early TNBC chemoresistance and thus need further confirmation in a larger and independent validation cohort. Citation Format: Nouritza Torossian, Dominika Foretek, Marc Gabriel, Linda Larbi Cherif, Charlotte Lecerf, Maud Kamal, Christophe Le Tourneau, Daniel Gautheret, Sergio Roman-Roman, Antonin Morillon. Reference free transcriptomic characterization of chemoresistant triple negative breast cancers provides a promising reservoir of predictive biomarkers of early chemoresistance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2774.