The Effect Of The Proportion Of Tumor Cells In Tumor Tissues On The Genomic Characteristics Analysis Of Tumor Subtypes

Primary tumor tissues are usually composed of two types of cells,one is tumor epithelial cells and the other one is stromal cells.And the latter include immune cells,fibroblasts,endothelial cells and normal epithelial cells etc.Due to the variability of tumor tissues,tumor tissues taken from surgical sampling are comprised of diverse proportions of tumor epithelial cells and stromal cells,ie the proportion of tumor cells(total number of tumor epithelial cells /(total number of tumor epithelial cells + total number of stromal cells))is different.Further,the proportion of tumor cells may have influence on cancer genome and transcriptome studies.To tackle this problem,some researchers propose genetic mutation calling algorithms(such as MuTect2)with parameters correcting the proportion of tumor cells.Others just developing new algorithms(such as ABSOLUTE)to evaluate the proportion of tumor cells directly.However,these algorithms often contain dozens of parameters,which need to be tuned carefully when applied to clinical samples,which brings great obstacles for clinical application.What’s more,the proportion of tumor cells computed by these algorithms is in low consistency with the proportion estimated by clinical immunohistochemistry,so the gold standard for evaluating the proportion of tumor cells in clinical is still based on the visual examination of pathological sections by pathologists.Using colon and breast cancer expression data from TCGA database,our previous transcriptomic analysis found that there is a significant correlation between the expression measurements of a large number of genes and the proportion of tumor cells,indicating that gene expression is affected by the proportion of tumor cells.Therefore,it is also necessary to assess the effect of the proportion of tumor cells on cancer genomic mutation analysis.Using the mutation data of 32 cancer types from the TCGA database and their corresponding proportion of tumor cells information,we analyzed the effect of the proportion of tumor cells on mutation detection.First,for each cancer,we evaluated the correlation between the number of gene mutations obtained by different mutation calling methods and the proportion of tumor cells.It was found that the weakest correlation algorithm was MuTect2,indicating that MuTect2 algorithm had a certain correction ability to the proportion of cancer cells compared with other algorithms.Taking the MuTect2 algorithm as an example,the results showed that there was a significant positive correlation between the number of gene mutations and the proportion of cancer cells in eight cancer types(gastric adenocarcinoma,breast cancer,lung squamous cell carcinoma,etc.)(Spearman rank correlation analysis,p<0.05),indicating that the number of gene mutations is affected by the proportion of tumor cells.Next,we selected some subtypes of seven common cancers,such as lung cancer,stomach adenocarcinoma and breast cancer,to analyze the effect of the proportion of cancer cells on the number of gene mutations among subtypes.For example,we found that the number of gene mutations in lung squamous cell carcinoma is significantly higher than that in lung adenocarcinoma,and the tumor cell composition ratio is also significantly higher than that in the lung adenocarcinoma.This implies that the distinction of the number of gene mutations in two subtypes may be caused by the difference in the proportion of cancer cells.Subsequently,we analyzed the relationship between the proportion of cancer cells and gene mutation rate among subtypes from multiple cancers.In the case of diffuse stomach adenocarcinoma,when the proportion of cancer cells was 70% or more,the mutation frequencies of 10 genes,such as MUC16,LRP1 B,FAT4,were all greater than 0.2558 in 43 samples;when the proportion of cancer cells was reduced to 60%-70%(including 60%),the mutation frequencies of the above 10 genes decreased in 18 samples;and when the proportion of cancer cells was less than 60%,the mutation frequencies of the above 10 genes in 8 samples decreased to 0.This indicates that when the proportion of cancer cells is relatively low,false negative results will appear in mutation detection of some genes,which may affect the analysis of genomic characteristics of the tumor subtype.In conclusion,this study shows that the proportion of cancer cells in cancer tissue can affect the reliability of gene mutation detection and the analysis of genomic characteristics of cancer subtypes,which may further affect our understanding of cancer biology and the implementation of precise treatment.Therefore,it is necessary to continue to develop algorithms to correct the impact of cancer proportion.At the same time,we should pay attention to controlling the proportion of tumor cells when interpreting the cancer genome information under the current situation.