作用：为四个着丝粒重复单元 unit_161、unit_172、unit_186、unit_330 构建共识序列：从每个基因组提取该单元的命中、用 MAFFT 比对、折叠成共识，再把 Gifu 与 MG20 的共识相互比对得到跨材料共识。
入口：run_unit_mafft_consensus.sh
输入：Lotus_GifuT2T_v1.0.fasta、Lotus_MG20T2T_v1.0.fasta；每个单元的 BED gifu_unit_<unit>.bed 与 mg20_unit_<unit>.bed（BED6）。
输出：separate_then_merge_consensus/ 下 01_seqs/<tag>_<unit>.fa、02_msa/<tag>_<unit>.aln.fa、03_consensus/<tag>_<unit>.cons.fa、03_dist/<tag>_<unit>.dist.tsv、04_merge2cons_msa/<unit>.Gifu_MG20.cons.{fa,aln.fa}、05_final/<unit>.final.cons.fa 与 <unit>.final.dist.tsv。
运行：bash run_unit_mafft_consensus.sh（须在含两个基因组 FASTA 与 gifu_/mg20_ 前缀 BED 的目录中运行；blast.sh 除末行外均为注释）；单次调用：python get_consensus.py -i aln.fa -o cons.fa -d dist.tsv -t 0.5 --include-gap --name <name>。
工具：MAFFT --thread 32 --auto；bedtools getfasta -name -s、jaccard、intersect、sort、merge；BLAST+ makeblastdb -dbtype nucl、blastn -task blastn-short -word_size 7 -evalue 1e-10（outfmt 6）；get_consensus.py 用 Biopython AlignIO，Levenshtein 可选（内置纯 Python 回退）；THRESH=0.5。
English
**Purpose**: Build consensus sequences for the four centromeric repeat units `unit_161`, `unit_172`, `unit_186` and `unit_330`: hits of each unit are extracted from each genome, aligned with MAFFT and collapsed into a consensus, and the Gifu and MG20 consensuses are then aligned to each other to give a cross-accession consensus.
**Entry point**: `run_unit_mafft_consensus.sh`
**Inputs**: `Lotus_GifuT2T_v1.0.fasta`, `Lotus_MG20T2T_v1.0.fasta`; the per-unit BEDs `gifu_unit_.bed` and `mg20_unit_.bed` (BED6).
**Outputs**: under `separate_then_merge_consensus/`: `01_seqs/_.fa`, `02_msa/_.aln.fa`, `03_consensus/_.cons.fa`, `03_dist/_.dist.tsv`, `04_merge2cons_msa/.Gifu_MG20.cons.{fa,aln.fa}`, `05_final/.final.cons.fa` and `.final.dist.tsv`.
**Run**: `bash run_unit_mafft_consensus.sh` (must run from a directory holding the two genome FASTAs and the `gifu_`/`mg20_`-prefixed BEDs; `blast.sh` is commented out except its last line); a single call: `python get_consensus.py -i aln.fa -o cons.fa -d dist.tsv -t 0.5 --include-gap --name `.
**Tools**: MAFFT `--thread 32 --auto`; bedtools `getfasta -name -s`, `jaccard`, `intersect`, `sort`, `merge`; BLAST+ `makeblastdb -dbtype nucl`, `blastn -task blastn-short -word_size 7 -evalue 1e-10` (`outfmt 6`); `get_consensus.py` uses Biopython `AlignIO` with `Levenshtein` optional (a pure-Python fallback is built in); `THRESH=0.5`.