摘要
To investigate the neoantigen-mediated immune elimination signal, Wu and colleagues developed ESCCF metric (1). Using this ESCCF metric, Wu and colleagues demonstrate the existence of neoantigen-mediated tumor elimination signal, and this conclusion directly contradicts with the Van den Eynden and colleagues study, which reports lack of neoantigen depletion signal in untreated cancer genome (2).The most important metric developed by the Wu and colleagues study is ESCCF, and the ESRNA metric questioned by Claeys and Van den Eynden (3) is a by-product of the Wu and colleagues study, and this ESRNA signal represents one of the mechanisms employed by cancer cells to escape immune surveillance, and the other immune escape mechanisms include antigen presentation pathway downregulation and upregulation of the expression of immune suppressive molecules, including PD-L1, CTLA4, etc.Claeys and Van den Eynden provide no evidence to argue against the ESCCF metric; however, they show evidence to question the ESRNA metric (3). Their evidence is based on simulations using The Cancer Genome Atlas (TCGA) or GTEx datasets; however, their simulation process is biased in the following aspects: (i) In their simulations, the expressions of all genes are fixed and do not change according to the mutation status, and this assumption is violated, because mutation can influence the expression of target genes. (ii) The mutations are classified into 96 types based on the nucleotide alterations and also the flanking nucleotides, and the IC50 values of different mutation types are not the same. The distribution of these mutation types is biased in their simulated datasets (Supplementary Fig. S1A). (iii) In actual TCGA dataset, mutations tend to happen more frequently in nonexpressed genes, while in their simulated TCGA dataset, mutations tend to happen more in expressed genes (Supplementary Fig. 1B). (iv) Their GTEx dataset has an additional problem, because the expressions of genes of all samples of the same tissue are the same in this dataset, disregarding the mutation status or sample status. The above-mentioned bias in their simulation process could explain why they detect ESRNA signal in simulated datasets.For ESRNA quantification, because different genes naturally have different expression levels, to compare the expression difference between immunogenic and immunogenic mutations is usually not a simple task. ESCCF quantification will not be similarly influenced as ESRNA, because the cancer cell fraction (CCF) is not influenced by gene expression or other known factors, and ESCCF represents a robust and accurate tool for immunoediting elimination signal quantification.As to the ESCCF signal, Claeys and Van den Eynden could not provide evidence to argue against it. And this ESCCF metric is directly comparable with the neoantigen depletion signal investigated in Van den Eynden and colleagues study (2). In Van den Eynden and colleagues study, HLA-binding mutation ratio (HBMR) was proposed as a metric to quantify neoantigen depletion signal; however, as reported in a preprint (bioRxiv 2020.05.11.089540), the design of this HBMR metric is problematic, and it is not unexpected that they could not detect neoantigen depletion signal using their HBMR method. On the basis of our additional unpublished data, this ESCCF signal exists even when using neoantigens predicted by HLA-II–related immunogenicity algorithms, suggesting the robustness of this ESCCF metric. In conclusion, the neoantigen enrichment scores developed in the Wu and colleagues study represent an innovative and robust approach to quantify immunoediting signal.See the original Letter to the Editor, p. 971No disclosures were reported.The authors thank ShanghaiTech University high performance computing public service platform for computing services. This work was supported by Natural science foundation of Shanghai (21ZR1442400 to X.-S. Liu).