作用：把 TRF 记录变成每条染色体的候选着丝粒单体序列并跨染色体连接：对每个重复单元做 canonical 化（旋转与反向互补）、选出最能解释该家族 k×L 长度结构的单体长度 L，再自比对单体得到染色体对连接数与连通分量。
入口：run.sh
输入：gifu_cent.bed、mg20_cent.bed（TRF 记录，单元序列在指定列）；gifu_chr_top_monomer.tsv、mg20_chr_top_monomer.tsv（chr、单元长度、单元序列、span）；cluster.py 用 self.sim90.tsv。
输出：run.sh 写 lotus_chr_top_monomer.fa、lotus_monomer.* BLAST 库、self.allvsall.tsv、self.noSelf.tsv、self.sim80.tsv、self.sim90.tsv、top_hit_per_query.tsv、chr_pair_links.tsv；canon_unit_and_length.py 写扩展 TSV 与 length_summary.tsv；pick_monomer_from_clusters.py 写 cluster_monomer.tsv；cluster.py 把簇打印到 stdout。
运行：bash run.sh；python3 canon_unit_and_length.py -i mg20_cent.bed -o mg20.units.canonical.tsv --summary mg20.length_summary.tsv --seq-col 9；python3 pick_monomer_from_clusters.py -i clusters.tsv -o cluster_monomer.tsv --tol 6 --max-k 6 --top 5；python3 cluster.py（run.sh 第 62 行以上为注释记录需单独运行；cluster.py 硬编码 self.sim90.tsv，须在本目录运行）。
工具：BLAST+ makeblastdb -dbtype nucl、blastn -task blastn-short -dust no -soft_masking false -word_size 7 -evalue 1e-10 -max_target_seqs 10000，outfmt 6；awk/sort 过滤 pident>=80 && len>=100（self.sim80.tsv）与 pident>=90 && len>=140（self.sim90.tsv）；三个 Python 脚本仅用标准库；#CSUB 队列 c01、88 槽。
English
**Purpose**: Turn TRF records into a per-chromosome set of candidate centromeric monomer sequences and link them across chromosomes: canonicalise each repeat unit (rotation and reverse complement), pick the monomer length L that best explains the family's k×L length structure, and self-align the monomers to build chromosome-pair link counts and connected components.
**Entry point**: `run.sh`
**Inputs**: `gifu_cent.bed`, `mg20_cent.bed` (TRF records, unit sequence in the configured column); `gifu_chr_top_monomer.tsv`, `mg20_chr_top_monomer.tsv` (chr, unit length, unit sequence, span); `self.sim90.tsv` for `cluster.py`.
**Outputs**: `run.sh` writes `lotus_chr_top_monomer.fa`, the `lotus_monomer.*` BLAST database, `self.allvsall.tsv`, `self.noSelf.tsv`, `self.sim80.tsv`, `self.sim90.tsv`, `top_hit_per_query.tsv` and `chr_pair_links.tsv`; `canon_unit_and_length.py` writes the extended TSV and `length_summary.tsv`; `pick_monomer_from_clusters.py` writes `cluster_monomer.tsv`; `cluster.py` prints clusters to stdout.
**Run**: `bash run.sh`; `python3 canon_unit_and_length.py -i mg20_cent.bed -o mg20.units.canonical.tsv --summary mg20.length_summary.tsv --seq-col 9`; `python3 pick_monomer_from_clusters.py -i clusters.tsv -o cluster_monomer.tsv --tol 6 --max-k 6 --top 5`; `python3 cluster.py` (everything above line 62 of `run.sh` is a commented record to be run separately; `cluster.py` hardcodes `self.sim90.tsv` and must run in this directory).
**Tools**: BLAST+ `makeblastdb -dbtype nucl`, `blastn -task blastn-short -dust no -soft_masking false -word_size 7 -evalue 1e-10 -max_target_seqs 10000` with `outfmt 6`; awk/sort filters `pident>=80 && len>=100` (`self.sim80.tsv`) and `pident>=90 && len>=140` (`self.sim90.tsv`); all three Python scripts are standard library only; `#CSUB` queue c01, 88 slots.