Generalizability of potential biomarkers of response to CTLA-4 and PD-1 blockade therapy in cancer
Authors
Bortone, D.Vensko, S.
Entwistle, S.
Cogdill, A.
Monette, A.
Najjar, Y.
Sweis, R.
Tschernia, N.
Wennerberg, E.
Bommareddy, P.
Haymaker, C.
Khan, U.
McGee, H.
Park, W.
Sater, H.
Spencer, C.
Ascierto, M.
Barsan, V.
Popat, V.
Valpione, Sara
Wells, D.
Thorsson, V.
Zappasodi, R.
Rudqvist, N.
Vincent, B.
Affiliation
University of North Carolina – Chapel Hill, Chapel Hill, NC, USAIssue Date
2020
Metadata
Show full item recordAbstract
Background Multiple genomics-based biomarkers of response to immune checkpoint inhibition have been reported or proposed, including tumor mutation/neoantigen frequency, PD-L1 expression, T cell receptor repertoire clonality, interferon gene signature expression, HLA expression, and others.1 Although genomics associations of response have been reported, the primary studies have used a variety of data generation and processing techniques. There is a need for data harmonization and assessment of generalizability of potential biomarkers across multiple datasets. Methods We acquired patient-level RNA sequencing FASTQ data files from 10 data sets reported in seven pan-cancer PD-1 and CTLA-4 immune checkpoint inhibition trials with matched clinical annotations.2–7 We applied a common bioinformatics workflow for quality control, mapping to reference (STAR), generating gene expression matrices (SALMON), T cell receptor repertoire inference (MiXCR), extraction of immune gene signatures and immune subtypes,8 and differential gene expression analysis (DESeq2). We analyzed i) immunogenomics features proposed as biomarkers, and ii) gene expression signatures built from each trial for association with overall survival across the set of trials using univariable Cox proportional hazards regression. In all, we assessed 9 total immunogenomics features/signatures. P-values were adjusted for multiple testing using the Benjamini-Hochberg method. Results Of the 9 immunogenomics features assessed, cytolytic activity score and expression of the Follicular Dendritic Cell Secreted Protein gene (FDCSP) were associated with survival in two of seven studies, respectively (adjusted p < 0.05) (figure 1). No proposed biomarkers were significantly associated with survival in more than two studies. The sets of genes significantly associated with clinical benefit across the studies were highly disjoint, with only three genes significant in three studies and thirteen genes significant in two studies (figure 2). No genes were significantly associated with clinical benefit in more than three of seven studiesCitation
Bortone D, Vensko S, Entwistle S, Cogdill A, Monette A, Najjar Y, et al. 75 Generalizability of potential biomarkers of response to CTLA-4 and PD-1 blockade therapy in cancer. Journal for ImmunoTherapy of Cancer. 2020;8(Suppl 3):A81-A2.Journal
Journal for Immunotherapy of CancerDOI
10.1136/jitc-2020-SITC2020.0075Additional Links
https://dx.doi.org/10.1136/jitc-2020-SITC2020.0075Type
Meetings and ProceedingsLanguage
enae974a485f413a2113503eed53cd6c53
10.1136/jitc-2020-SITC2020.0075