注：腳本是用GPT生成的，請自己甄別。

1. perl gff3_to_bed12.pl in.gff > out.bed12

cat gff3_to_bed12.pl

#!/usr/bin/perl
use strict;
use warnings;

# 用法：
#   perl gff3_to_bed12.pl in.gff > out.bed12
#
# 要點(diǎn)：
# - 以 mRNA 的 ID 作為轉(zhuǎn)錄本鍵與 BED12 的 name（若有 Name 則優(yōu)先）。
# - exon/CDS 通過 Parent 歸并到對應(yīng) mRNA。
# - 坐標(biāo)：GFF 1-based 閉區(qū)間 -> BED 0-based 半開（start-1, end 原樣）。
# - 無 CDS 的轉(zhuǎn)錄本：thickStart=thickEnd=txStart（顯示為非編碼）。
# - blockSizes 與 blockStarts 不帶尾隨逗號。

my $gff = shift or die "Usage: $0 input.gff3 > output.bed12\n";
open my $IN, "<", $gff or die "Cannot open $gff: $!\n";

# 以轉(zhuǎn)錄本 ID 為鍵
my (%CHR, %STR, %NAME, %EXONS, %CDS_S, %CDS_E);

while (<$IN>) {
    chomp;
    next if /^#/;
    my ($chr, undef, $type, $start, $end, undef, $strand, undef, $attr) = split(/\t/);
    $type = lc $type;

    # 解析屬性為哈希（split 限制 2，兼容像 Class== 之類的情況）
    my %A;
    for my $kv (split /;/, $attr) {
        next unless length $kv;
        my ($k, $v) = split(/=/, $kv, 2);
        next unless defined $k;
        $A{$k} = defined $v ? $v : '';
    }

    # 轉(zhuǎn)換為 0-based 半開
    my $s0 = $start - 1;
    my $e0 = $end;

    # mRNA 或 transcript 行：記錄基礎(chǔ)信息
    if ($type eq 'mrna' or $type eq 'transcript') {
        my $tid = $A{ID} // next;
        $CHR{$tid}  = $chr;
        $STR{$tid}  = $strand;
        $NAME{$tid} = $A{Name} // $tid;
    }
    # exon/CDS：按 Parent 歸并
    elsif ($type eq 'exon' or $type eq 'cds') {
        my $parent = $A{Parent} // next;
        for my $tid (split /,/, $parent) {
            $tid =~ s/^\s+|\s+$//g;
            # 若 mRNA 行在后面出現(xiàn)，先占位
            $CHR{$tid}  //= $chr;
            $STR{$tid}  //= $strand;
            $NAME{$tid} //= $tid;

            if ($type eq 'exon') {
                push @{$EXONS{$tid}}, [$s0, $e0];
            } else { # cds
                $CDS_S{$tid} = $s0 if !defined $CDS_S{$tid} || $s0 < $CDS_S{$tid};
                $CDS_E{$tid} = $e0 if !defined $CDS_E{$tid} || $e0 > $CDS_E{$tid};
            }
        }
    }
}
close $IN;

# 輸出 BED12
for my $tid (keys %EXONS) {
    my @ex = sort { $a->[0] <=> $b->[0] } @{$EXONS{$tid}};
    next unless @ex;  # 必須有外顯子

    my $txS = $ex[0]->[0];
    my $txE = $ex[-1]->[1];

    my @blockSizes  = map { $_->[1] - $_->[0] } @ex;
    my @blockStarts = map { $_->[0] - $txS }    @ex;

    # CDS 范圍（無 CDS 則設(shè)為非編碼）
    my ($thickS, $thickE) = (defined $CDS_S{$tid} && defined $CDS_E{$tid})
        ? ($CDS_S{$tid}, $CDS_E{$tid})
        : ($txS, $txS);

    my $chrom  = $CHR{$tid} // 'chrN';
    my $strand = $STR{$tid} // '+';
    my $name   = $NAME{$tid} // $tid;

    my $sizes  = join(",", @blockSizes);   # 無尾逗號
    my $starts = join(",", @blockStarts);  # 無尾逗號

    print join("\t",
        $chrom, $txS, $txE, $name, 0, $strand,
        $thickS, $thickE, 0,
        scalar(@ex),
        $sizes,
        $starts
    ), "\n";
}