跟着PNAS学数据分析：泛基因组（pan-genome）分析核心基因组可变基因组大小

论文

Novel functional sequences uncovered through a bovine multiassembly graph

代码链接

https://github.com/AnimalGenomicsETH/bovine-graphs/tree/main

代码是一个snakemake流程，慢慢看，争取把每个rule都拆解出来

基本思路

使用minigraph做基因组比对，获得一个图基因组，图基因组包含边和节点，节点是序列。然后把每个基因组单独比对回图基因组，可以判断图基因组中节点是否被覆盖，如果所有基因组都覆盖这个节点，这个节点就是核心基因组的一部分，否则就是可变基因组

这里需要理解一下gfa格式的文件

论文提供了分析流程用到的代码，我们用拟南芥的数据试试，拟南芥的论文

Chromosome-level assemblies of multiple Arabidopsis genomes reveal hotspots of rearrangements with altered evolutionary dynamics

https://www.nature.com/articles/s41467-020-14779-y

这个论文里就做了核心基因组和可变基因组的分析，但是这里的方法和PNAS牛的这篇文章不一样

代码

7个拟南芥基因组序列，只用组装到染色体水平的序列

代码语言：javascript复制

seqkit grep -r -f chr.list ../../An-1.chr.all.v2.0.fasta -o 00.assembly/An1.fa
seqkit grep -r -f chr.list ../../C24.chr.all.v2.0.fasta -o 00.assembly/C24.fa
seqkit grep -r -f chr.list ../../Cvi.chr.all.v2.0.fasta -o 00.assembly/Cvi.fa
seqkit grep -r -f chr.list ../../Eri.chr.all.v2.0.fasta -o 00.assembly/Eri.fa
seqkit grep -r -f chr.list ../../Kyo.chr.all.v2.0.fasta -o 00.assembly/Kyo.fa
seqkit grep -r -f chr.list ../../Ler.chr.all.v2.0.fasta -o 00.assembly/Ler.fa
seqkit grep -r -f chr.list ../../Sha.chr.all.v2.0.fasta -o 00.assembly/Sha.fa

minigraph构建图基因组，参考基因组放第一个位置

代码语言：javascript复制

time minigraph --inv no -xggs -t 48 
00.assembly/An1.fa 00.assembly/C24.fa 
00.assembly/Cvi.fa 00.assembly/Eri.fa 
00.assembly/Kyo.fa 
00.assembly/Ler.fa 
00.assembly/Sha.fa > at.gfa

8m50.144s

minigraph再把每个基因组比对到图基因组上

这里gaf和gfa有啥区别？

代码语言：javascript复制

minigraph -t 48 --cov -x asm at.gfa 00.assembly/An1.fa > An1.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/C24.fa > C24.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/Cvi.fa > Cvi.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/Eri.fa > Eri.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/Kyo.fa > Kyo.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/Ler.fa > Ler.gaf
minigraph -t 48 --cov -x asm at.gfa 00.assembly/Sha.fa > Sha.gaf

从gaf文件中解析每个node的覆盖信息

代码语言：javascript复制

python comb_coverage01.py -g An1.gaf -a An1 -o An1Cov.tsv -r Y
python comb_coverage01.py -g C24.gaf -a C24 -o C24Cov.tsv -r N
python comb_coverage01.py -g Cvi.gaf -a Cvi -o CviCov.tsv -r N
python comb_coverage01.py -g Eri.gaf -a Eri -o EriCov.tsv -r N
python comb_coverage01.py -g Ler.gaf -a Ler -o LerCov.tsv -r N
python comb_coverage01.py -g Sha.gaf -a Sha -o ShaCov.tsv -r N

合并数据

代码语言：javascript复制

library(tidyverse)

read_tsv("D:/Jupyter/PNAS_bovine/An1Cov.tsv") %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/C24Cov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/CviCov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/EriCov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/KyoCov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/LerCov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  left_join(read_tsv("D:/Jupyter/PNAS_bovine/ShaCov.tsv"),
            by=c("nodeid"="nodeid")) %>% 
  select(-c(2:5)) -> datmat


datmat %>% 
  column_to_rownames("nodeid") -> datmat

datmat[datmat>0] <- 1
datmat %>% 
  rownames_to_column("nodeid") %>% 
  write_tsv("D:/Jupyter/PNAS_bovine/nodemat.tsv")

image.png

代码语言：javascript复制

awk '$1~/S/ {{ split($5,chr,":"); split($6,pos,":"); split($7,arr,":");print $2,length($3),chr[3],pos[3],arr[3] }}' at.gfa > graph_len.tsv

这段awk的代码不是太明白，对比输出和输入能看出来是在做啥

核心基因组大小

代码语言：javascript复制

nodemat<-"D:/Jupyter/PNAS_bovine/nodemat.tsv"
datmat <- read.table(nodemat, header = TRUE, stringsAsFactors = FALSE)

datmat$combres <- rowSums(datmat %>% select(-nodeid))
datmat
totassemb <- ncol(datmat) - 2
totassemb

graphlen<-"D:/Jupyter/PNAS_bovine/graph_len.tsv"
datlen <- read.table(graphlen, header = FALSE, stringsAsFactors = FALSE)
colnames(datlen) <- c("nodeid", "conlen", "chromo", "pos", "rrank")
datlen
datmat <- datmat %>% left_join(datlen, by = c("nodeid"))
datmat %>% head()

core <- sum(datmat[datmat$combres == totassemb, "conlen"])
core

这个是95M,论文中写的是105M

泛基因组曲线

代码语言：javascript复制

datexp <- read.table(nodemat, header = TRUE, stringsAsFactors = FALSE)
datexp
datlen <- read.table(graphlen, header = FALSE, stringsAsFactors = FALSE)
datlen %>% head()

colnames(datlen) <- c("nodeid", "conlen", "chromo", "pos", "rrank")


breeds <- colnames(datexp %>% select(-nodeid))
breeds
no_rep <- ncol(datexp) - 1
no_rep


datpan <- data.frame(
  norep = numeric(),
  nosamp = numeric(),
  assemb = character(),
  core_gen = numeric(),
  flex_gen = numeric(),
  tot_gen = numeric()
)

datcon <- datlen %>% select(nodeid, conlen)
datcon


for (nosamp in seq(1, length(breeds))) {
  # how many sampling repeated
  for (norep in seq(1, no_rep)) {
    selsamp <- as.character(sample(breeds, size = nosamp))
    # selected breeds
    selcol <- datexp[, colnames(datexp) %in% selsamp, drop = FALSE]
    # add contig length
    # it is ordered so we just add it from conlen
    # add number of colour in node
    selcol$comcol <- rowSums(selcol)
    # add contig len
    selcol$conlen <- datlen$conlen
    # core genome
    # core if shared in all of the member of population
    core_gen <- selcol[selcol$comcol == nosamp, "conlen"] %>% sum()
    # flex genome if shared less than the total of population
    # not consider node not present 
    if (nosamp == 1) 
      flex_gen  <- 0 
    else 
      flex_gen <- selcol[selcol$comcol < nosamp & selcol$comcol > 0, "conlen"] %>% sum()
    # tot_genome if at least observed in a single breeds
    tot_gen <- selcol[selcol$comcol > 0, "conlen"] %>% sum()
    datpan <- rbind(datpan, data.frame(
      norep = norep,
      nosamp = nosamp,
      assemb = paste(selsamp, collapse = ","),
      core_gen = core_gen,
      flex_gen = flex_gen,
      tot_gen = tot_gen
    ))
  }
}

datpan

代码语言：javascript复制

p1<-datpan %>% 
  select(nosamp,core_gen,tot_gen) %>% 
  mutate(nosamp=factor(nosamp)) %>% 
  pivot_longer(!nosamp) %>% 
  ggplot(aes(x=nosamp,y=value)) 
  geom_boxplot(aes(fill=name)) 
  geom_smooth(aes(x=as.numeric(nosamp),color=name)) 
  #geom_violin(aes(fill=name)) 
  theme_bw(base_size = 20) 
  theme(panel.grid = element_blank(),
        legend.position = "top",
        legend.title = element_blank()) 
  scale_y_continuous(labels = function(x){paste(x/1000000,"M")}) 
  labs(x="Sample Number",y="Genome Size") 
  scale_color_manual(values = c("#e3010a","#00b2ec")) 
  scale_fill_manual(values = c("#e3010a","#00b2ec"))

p2<-datpan %>% 
  select(nosamp,core_gen,tot_gen) %>% 
  mutate(nosamp=factor(nosamp)) %>% 
  pivot_longer(!nosamp) %>% 
  ggplot(aes(x=nosamp,y=value,color=name)) 
  geom_jitter(width = 0.1,size=5,alpha=0.8) 
  geom_smooth(aes(x=as.numeric(nosamp))) 
  #geom_violin(aes(fill=name)) 
  theme_bw(base_size = 20) 
  theme(panel.grid = element_blank(),
        legend.position = "top",
        legend.title = element_blank()) 
  scale_y_continuous(labels = function(x){paste(x/1000000,"M")}) 
  labs(x="Sample Number",y="Genome Size") 
  scale_fill_manual(values = c("#4da0a0","#9b3a74")) 
  scale_color_manual(values=c("#4da0a0","#9b3a74"))

library(patchwork)
p1 p2

image.png

数据分析 assembly core 论文 数据

1 人点赞