1.背景知识
在一篇文章里看到了微卫星不稳定性（Microsatellite Instability，MSI）与riskscore的关系图，就去查了一下，很好的背景知识资料：
https://mp.weixin.qq.com/s/zx5eaNvBrwWxKWgIl7CC4w
核心知识：
1.计算MSI分数的工具：MANTIS，默认阈值0.4，高于阈值为MSI-H，低于阈值为MSS(无明显的MSI出现)。
2.最早再结直肠癌种发现，是预后良好的标志，MSI结直肠癌5年生存率要显著高于MSS结直肠癌，MSI-H结直肠癌比MSS结直肠癌有更好的预后。
2.寻找TCGA的MSI数据
那么TCGA微卫星不稳定的数据上哪找呢？隐约记得在cbioportal有。但是他们现在把下载接口删掉了，不好找。
继续搜索，搜到了这个包：cBioPortalData，可以下载数据。
一个不错的教程：
https://zhuanlan.zhihu.com/p/406088178
我的需求，主要是下载cBioPortal的临床信息，有很多是其他渠道找不到的。
使用起来非常简单：
library(cBioPortalData)
cbio <- cBioPortal()
studies = getStudies(cbio)
head(studies$studyId)
## [1] "acc_tcga" "blca_plasmacytoid_mskcc_2016"
## [3] "bcc_unige_2016" "all_stjude_2015"
## [5] "ampca_bcm_2016" "blca_dfarber_mskcc_2014"
id = "ucec_tcga_pan_can_atlas_2018"
clinical = clinicalData(cbio, id)
colnames(clinical)
## [1] "patientId"
## [2] "AGE"
## [3] "AJCC_STAGING_EDITION"
## [4] "BUFFA_HYPOXIA_SCORE"
## [5] "CANCER_TYPE_ACRONYM"
## [6] "DAYS_LAST_FOLLOWUP"
## [7] "DAYS_TO_INITIAL_PATHOLOGIC_DIAGNOSIS"
## [8] "DSS_MONTHS"
## [9] "DSS_STATUS"
## [10] "ETHNICITY"
## [11] "FORM_COMPLETION_DATE"
## [12] "HISTORY_NEOADJUVANT_TRTYN"
## [13] "ICD_10"
## [14] "ICD_O_3_HISTOLOGY"
## [15] "ICD_O_3_SITE"
## [16] "INFORMED_CONSENT_VERIFIED"
## [17] "IN_PANCANPATHWAYS_FREEZE"
## [18] "NEW_TUMOR_EVENT_AFTER_INITIAL_TREATMENT"
## [19] "OS_MONTHS"
## [20] "OS_STATUS"
## [21] "OTHER_PATIENT_ID"
## [22] "PERSON_NEOPLASM_CANCER_STATUS"
## [23] "PFS_MONTHS"
## [24] "PFS_STATUS"
## [25] "PRIOR_DX"
## [26] "RACE"
## [27] "RADIATION_THERAPY"
## [28] "RAGNUM_HYPOXIA_SCORE"
## [29] "SAMPLE_COUNT"
## [30] "SEX"
## [31] "SUBTYPE"
## [32] "WEIGHT"
## [33] "WINTER_HYPOXIA_SCORE"
## [34] "DAYS_TO_BIRTH"
## [35] "DFS_MONTHS"
## [36] "DFS_STATUS"
## [37] "sampleId"
## [38] "ANEUPLOIDY_SCORE"
## [39] "CANCER_TYPE"
## [40] "CANCER_TYPE_DETAILED"
## [41] "FRACTION_GENOME_ALTERED"
## [42] "GRADE"
## [43] "MSI_SCORE_MANTIS"
## [44] "MSI_SENSOR_SCORE"
## [45] "MUTATION_COUNT"
## [46] "ONCOTREE_CODE"
## [47] "SAMPLE_TYPE"
## [48] "SOMATIC_STATUS"
## [49] "TISSUE_PROSPECTIVE_COLLECTION_INDICATOR"
## [50] "TISSUE_RETROSPECTIVE_COLLECTION_INDICATOR"
## [51] "TISSUE_SOURCE_SITE"
## [52] "TISSUE_SOURCE_SITE_CODE"
## [53] "TMB_NONSYNONYMOUS"
## [54] "TUMOR_TISSUE_SITE"
## [55] "TUMOR_TYPE"
这是临床信息的数据列名，里面就包括了MSI_SCORE_MANTIS这一列。
3.画个图玩
R语言的好处就是拿到了数据可以进行自定义的可视化，比网页工具更加灵活，也更好重复。
df = na.omit(clinical[,c("patientId","MSI_SCORE_MANTIS")])
colnames(df)[2] = "MSI_score"
df$MSI_score = as.numeric(df$MSI_score)
k= df$MSI_score >0.4;table(k)
## k
## FALSE TRUE
## 358 168
df$Group = ifelse(k,"MSI","MSS")
head(df)
## # A tibble: 6 x 3
## patientId MSI_score Group
## <chr> <dbl> <chr>
## 1 TCGA-2E-A9G8 0.323 MSS
## 2 TCGA-4E-A92E 0.340 MSS
## 3 TCGA-5B-A90C 0.334 MSS
## 4 TCGA-5S-A9Q8 0.320 MSS
## 5 TCGA-A5-A0G1 0.311 MSS
## 6 TCGA-A5-A0G2 0.400 MSI
整理好了数据，就可以画图啦
library(ggplot2)
ggplot(df,aes(x = Group,y = MSI_score,fill = Group))+
geom_boxplot()+
geom_jitter(size = 0.5)+
geom_hline(yintercept = 0.4,lty = 4)+
theme_bw()
网友评论