單細(xì)胞轉(zhuǎn)錄組探索CRC病人的一致性
文章是:Reference component analysis of single-cell transcriptomes elucidates cellular heterogeneity in human colorectal tumors. Nat Genet 2017 May;49(5):708-718. PMID: 28319088
單細(xì)胞轉(zhuǎn)錄組笼沥,使用的是GPL11154Illumina HiSeq 2000 (Homo sapiens) ,數(shù)據(jù)都上傳到了:GSE81861
BioProject | PRJNA323703 |
---|---|
SRA | ERP016958 |
既有病人的單細(xì)胞轉(zhuǎn)錄組數(shù)據(jù),同時(shí)也有細(xì)胞系的數(shù)據(jù)做驗(yàn)證。
- 1,591 single cells from 11 colorectal cancer patients,包括 969腫瘤部位細(xì)胞以及 622 癌旁細(xì)胞。嚴(yán)格過(guò)濾后只剩下:375 tumor cells and 215 normal mucosa cells
- 630 single cells from 7 cell lines,過(guò)濾后剩下561個(gè)
- 83 A549 cells,
- 65 H1437 cells,
- 55 HCT116 cells,
- 23 IMR90 cells,
- 96 K562 cells,
- 134 GM12878 cells (38 from batch 1, 96 from batch 2)
- 174 H1 cells (96 from batch 1, 78 from batch 2).
上游測(cè)序數(shù)據(jù)沒(méi)有必要重新下載分析了蝠筑,可以直接使用作者上傳的表達(dá)矩陣:
Supplementary file | Size | Download | File type/resource |
---|---|---|---|
GSE81861_CRC_NM_all_cells_COUNT.csv.gz | 3.2 Mb | (ftp)(http) | CSV |
GSE81861_CRC_NM_all_cells_FPKM.csv.gz | 4.7 Mb | (ftp)(http) | CSV |
GSE81861_CRC_NM_epithelial_cells_COUNT.csv.gz | 2.5 Mb | (ftp)(http) | CSV |
GSE81861_CRC_NM_epithelial_cells_FPKM.csv.gz | 4.0 Mb | (ftp)(http) | CSV |
GSE81861_CRC_tumor_all_cells_COUNT.csv.gz | 4.3 Mb | (ftp)(http) | CSV |
GSE81861_CRC_tumor_all_cells_FPKM.csv.gz | 7.9 Mb | (ftp)(http) | CSV |
GSE81861_CRC_tumor_epithelial_cells_COUNT.csv.gz | 3.6 Mb | (ftp)(http) | CSV |
GSE81861_CRC_tumor_epithelial_cells_FPKM.csv.gz | 6.5 Mb | (ftp)(http) | CSV |
GSE81861_Cell_Line_COUNT.csv.gz | 13.1 Mb | (ftp)(http) | CSV |
GSE81861_Cell_Line_FPKM.csv.gz | 28.9 Mb | (ftp)(http) | CSV |
GSE81861_GEO_EGA_ID_match.csv.gz | 14.4 Kb | (ftp)(http) | CSV |
作者認(rèn)為全文最重要的是開(kāi)發(fā)了一個(gè)挖掘細(xì)胞類型的算法:reference component analysis (RCA) 優(yōu)于其它現(xiàn)有的算法】粒可以把cancer-associated fibroblasts (CAFs)繼續(xù)分成兩個(gè)類別什乙。對(duì)比的算法包括:
- hierarchical clustering using all expressed genes (All-HC)
- hierarchical clustering using principal-component analysis (PCA)-based feature selection (HiLoadG-HC)
- BackSPIN
- RaceID2
- Seurat
- three additional methods based on selection of genes with highly variable expression (VarG-HC, VarG-PCAproj-HC and VarG-tSNEproj-HC).
使用 adjusted Rand index (ARI) 指標(biāo)來(lái)評(píng)價(jià)各個(gè)聚類算法的優(yōu)劣。結(jié)果發(fā)現(xiàn)自己開(kāi)發(fā)的RCA表現(xiàn)超常R亚颉N惹俊!
當(dāng)然了和悦,還在 Tirosh, I. et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA–seq. Science 352, 189–196 (2016). 文章的數(shù)據(jù)里面做了驗(yàn)證退疫。
背景知識(shí)
腫瘤異質(zhì)性很重要,單細(xì)胞轉(zhuǎn)錄組測(cè)序很厲害鸽素,以前的研究根據(jù)單細(xì)胞轉(zhuǎn)錄組表達(dá)矩陣進(jìn)行分類的算法不夠好褒繁,所以他們開(kāi)發(fā)reference component analysis (RCA) , 而且 Colorectal cancer (CRC) 疾病非常嚴(yán)重馍忽,需要探索棒坏。
根據(jù)細(xì)胞系單細(xì)胞表達(dá)數(shù)據(jù)探索算法
630個(gè)細(xì)胞的表達(dá)數(shù)據(jù),過(guò)濾后剩下561個(gè)遭笋,這里使用Fragments per kilobase per million reads (FPKM)來(lái)進(jìn)行表達(dá)定量坝冕。因?yàn)槠渖嫌翁幚碜叩氖荰OPHAT2+CUFFLINKS流程。
單細(xì)胞過(guò)濾策略
rate of exonic reads (ROER) 需要大于5%
number of detected genes (NODG) 需要大于1000瓦呼, 基因的FPKM ≥1才能算被檢測(cè)到了喂窟。
Exonic reads (ER) 要大于0.1Million
管家基因: TFRC, ACTB, RPLP0, PGK1, GAPDH, LDHA, NONO, B2M, GUSB and PPIH.
RCA算法細(xì)節(jié)
首先從 BioGPS數(shù)據(jù)庫(kù)里面下載兩個(gè)數(shù)據(jù)集: HumanU133A/GNF1H Gene Atlas and the Primary Cell Atlas ,從中挑選 A total of 4,717 genes were selected as features for GNF1H and 5,209 genes were selected for the Primary Cell Atlas. 還使用了 WGCNA 算法央串。
還使用了一些其它公共數(shù)據(jù):TCGA, GSE14333, the PRECOG database, and GSE33113, GSE37892 and GSE39582 來(lái)驗(yàn)證單細(xì)胞轉(zhuǎn)錄組得到的基因集(The 'fibroblast-like' signature )是否能顯著的區(qū)分CRC病人的生存情況磨澡。
需要了解一些細(xì)胞類型的 known markers
- epithelial cells (VIL1, KRT20, CLDN7, CDH1)
- endothelial cells (Endo; ENG)
- fibroblasts (Fibro; SPARC, COL14A, COL3A1, DCN)
- B cells (CD38, MZB1, DERL3)
- T cells (TRBC2, CD3D, CD3E, CD3G)
- myeloid cells (ITGAX, CD68, CD14, CCL3)
- mast cells (KIT, TPSB2)
做成了一個(gè)R包供使用:RCA R package, https://github.com/GIS-SP-Group/RCA.
(文章轉(zhuǎn)自jimmy的2018年閱讀文獻(xiàn)筆記)
生信基礎(chǔ)知識(shí)大全系列:生信基礎(chǔ)知識(shí)100講
史上最強(qiáng)的生信自學(xué)環(huán)境準(zhǔn)備課來(lái)啦!质和! 7次改版稳摄,11節(jié)課程,14K的講稿饲宿,30個(gè)夜晚打磨厦酬,100頁(yè)P(yáng)PT的課程胆描。
如果需要組裝自己的服務(wù)器;代辦生物信息學(xué)服務(wù)器
如果需要幫忙下載海外數(shù)據(jù)(GEO/TCGA/GTEx等等)仗阅,點(diǎn)我袄友?
如果需要線下輔導(dǎo)及培訓(xùn),看招學(xué)徒
如果需要個(gè)人電腦:個(gè)人計(jì)算機(jī)推薦
如果需要置辦生物信息學(xué)書(shū)籍霹菊,看:生信人必備書(shū)單
如果需要實(shí)習(xí)崗位:實(shí)習(xí)職位發(fā)布
如果需要售后:點(diǎn)我
如果需要入門資料大全:點(diǎn)我