Caveolin-1 and Sox-2 are Predictive Biomarkers of Cetuximab Response in Head and Neck Cancer

Expression analysis (RNA-seq) and RPPA data (see publication) of head and neck cancer patient-derived xenographs (PDXs) revealed Sox-2 and Cav-1 expression to be biomarker determinants of head and neck cancer patient tumor response to cetuximab (EGFR inhibitor) treatment. Overall design: Total RNA was extracted from 69 whole PDX tumors using the RNAeasy kit (Qiagen) following manufacturer's instructions (only 65 were retained for downstream analyses). Total RNA was enriched for poly-A transcripts and sequenced using the NovaSeq 6000 platform (100PE) at the QB3-Berkeley Genomics center at UC Berkeley. Following sequencing, the unaligned reads contained a mix of human stroma, mouse germline, viral and ostensibly other sequences. Hence, prior to mapping to human reference and expression counting, reads were classified to a probable source. For convenience, each sample had been previously split between two fastq files. Adapters were trimmed and reads failing QC removed using fastp (v0.20.0). Reads were classified as "human," "mouse," "both," "neither" and "ambiguous" using Xenome (v1.0.0), a k-mer index based pseudoaligner, with each sample's split files classified independently of each other. Only reads classified as uniquely human progressed to the human expression pipeline. Uniquely human reads were merged into a single fastq file for each sample and next aligned against a ribosomal reference using bwa, removing rRNA matches. Finally, remaining reads were mapped against an Ensembl GRCh38.90 reference. Expression counts were estimated using RSEM (v1.3.1) (see processed TPM file).