问题背景

大致是这样，假如你用一堆注释软件然后用不同物种得到了一堆不同的gff，那这些gff肯定是有重叠的，对于注释信息gff来说只需要保留gene、exon、cds有这些features的区段，并且去除有重叠的区段仅仅保留一个，这里不考虑打分长度什么的（不同软件不同物种也不好比较分数，当然有些软件可以根据权重来合并注释结果，比如RNA注释权重较高然后同源注释次之从头预测的最低），于是脚本就如下（其实用python/R也很简单，但是考虑到性能方面还是用zsh/以awk为主体会快很多），具体思路就是以features为gene的行作为标识符（因为这个gene feature下面所有的记录都是针对这个基因的，所以相当于分隔记录符一样），用sort排序得到对应标识符顺序，用awk读进去成数组，生成“标识符：针对这个gene的所有记录”的哈希表，然后根据排序的数组的标识符顺序来得到整个gff的排序。接下来再判断排序后的gff的"gene features"的首尾区间是否有重叠，同样也是用哈希记录然后最终输出。(简书这个缩进我属实无语懒得改了)

#!/usr/bin/zsh
# -*- coding: utf-8 -*-
### ------------------------------------
### merge gff results generated by annotation pipeline
### ------------------------------------

# Get gffdir from input
while {getopts d: arg} {
        case $arg {
                (d)
                gffdir=$OPTARG
                d=$arg
                # Test if the gffdir exists.
                if [[ -d $gffdir ]] {
                        echo "Your gffdir is $gffdir"
                } else {
                           echo "The directory that you specified does not exist, please specify the correct path."
                           exit
                }
                ;;
                (?)
                echo "Wrong option!"
                ;;
        }
}

#  Test if the -d option is provided.
if [[ -z $d ]] {
        echo "You must use -d to specify the directory that contains all the gff files that you want to merge."
        exit
}

# Merge all gff to a big gff
if [[ -f $gffdir/merged.gff ]] {
        rm -rf $gffdir/merged.gff
}
if [[ -f $gffdir/sortedmerged.gff ]] {
        rm -rf $gffdir/sortedmerged.gff
}
if [[ -f $gffdir/filtermerged.gff ]] {
        rm -rf $gffdir/filtermerged.gff
}
ls $gffdir/*gff | while read gff
do
        print "Proccessing $gff..."
        grep -v "^#" $gff >> $gffdir/merged.gff
done && print "Successfully merged gff files!"

# sort gff
## Get a sorted id array that contains only the rows whose feature is gene
id_sorted=`gawk '$3=="gene"{print $0}' $gffdir/merged.gff |
        sort -t $'\t' -k 1,1 -k 7,7 -k 4n,4 -k 5n,5`
## Use awk and srotedIDarray to get a sorted gff file
gawk -v arr=$id_sorted '
BEGIN{
split(arr,id_sorted,"\n")
RS="\n"
FS="\t"
OFS="\t"
ORS="\n"
}
$3=="gene"{
        recs[id]=id"\n"lines
        id=$0
        lines=""
}
$3!="gene"{
        lines=lines"\n"$0
}
END{
recs[id]=id+"\n"+lines
for (id in id_sorted){
        # print recs[id]
        print recs[id_sorted[id]]
}
}
' $gffdir/merged.gff | sed '/^\s*$/d' > $gffdir/sortedmerged.gff


# Exclude overlap records of sorted gff
gawk 'BEGIN{
RS="\n"
FS="\t"
flag=1
}
$3=="gene"{
if(!start[$1","$7]){
        start[$1","$7]=$4
        end[$1","$7]=$5
        flag=1
        print $0
}
else{
        s=start[$1","$7]
        e=end[$1","$7]
        if((s<=$4&&$4<=e)||(s<=$5&&$5<=$e)){
                flag=0
        }
        else{
                start[$1","$7]=$4
                end[$1","$7]=$5
                flag=1
                print $0
        }
}
}
$3!="gene"&&flag==1{print $0}
' $gffdir/sortedmerged.gff > $gffdir/filtermerged.gff