08 单核转录组
作用:把 Gifu T2T 基因组上的单核 RNA-seq(根的两个阶段 Dpi00 与 Dpi02,各两个重复)从 Cell Ranger 定量做到聚类与细胞类型注释、簇/细胞类型 DEG 与 scTenifoldKnk 虚拟敲除。
入口:下表 7 个 shell 脚本;Script/*.py 中需手动运行的为 03_1-cluster_statitcs.py、03_2-fliter_to_all_gene_h5ad.py、05_3-marker_cluster_DEGs_own_order.py 与 makesure_02dpi_samples_all_have_Early_signal_cell.py。
输入:sample.list(ProjectID CNName RunID SampleName)、四个 *_filtered.h5ad、Gifu T2T FASTA/GTF 与 .pep.fa、{Gmax,Gm2.1,GmZH,At}_marker_genes.pep.fa、eggNOG 注释;后续步骤按当前目录读 02_processData/ 与 04_Anno/,scTenifoldKnk 另需 zz_00_pea.rds 与 All_species_merged_by_GifuT2T.SNF_Symbol.tsv。
输出:00_ref/、01_cellranger_out/<SampleName>/、02_processData/(h5ad、zz_*、zz_00_pea.rds、00_02_pseudobulk_fpkm.csv)、03_figures/(降维图、clustree_plot_*.pdf、火山图)、04_Anno/(Seurat rds、marker CSV、blast_*/、富集结果)、06_batch_multiple_scTenifoldKnk/<group>/<gene>/。
运行:在本阶段根目录(含 02_processData/、03_figures/、04_Anno/)按表中顺序运行,编号脚本携带 #CSUB 作业指令、交由调度器提交;{TOKEN} 路径占位符见 scripts/README.md;注意 03_1-dimenRedu_adjusted.py 写出的文件名与 03_1-cluster_statitcs.py、03_2-fliter_to_all_gene_h5ad.py 读取的文件名不一致,需先统一。
工具:Cell Ranger 9.0.1;Python 3.9.19 + scanpy/anndata/scvi-tools/leidenalg/rpy2;R 4.4.1 + Seurat/scTenifoldKnk/DESeq2/clustree;BLAST+ 与 seqkit;参数:n_top_genes 4000(扫描 1000–5000)、Leiden 0.4(扫描 0.2–1.2)、每基因 100 次 scTenifoldKnk 重复(K ∈ {10,30,50,70,100})。
| 脚本 | 用途 |
|---|---|
00_mkref.sh |
cellranger mkref --genome=Gifu(Gifu T2T FASTA + GTF)→ ./00_ref |
00_cellranger.sh |
对 sample.list 中每个样本执行一次 cellranger count → 01_cellranger_out/<SampleName> |
run_singlecell.sh |
一个驱动依次跑 Script/01_annaDataQC.py、02_integrate.py、03_1-dimenRedu_adjusted.py、05_3-markerAnno_Ps_v5_own.py、06_4-DEGs_cell_type.R |
05_cluster_DEGs_finder.sh |
先 05_1-…py 再 05_2-…R:Seurat 对象与簇 marker,随后按簇做富集 |
06_1.marker_gene_finder.sh |
DEG → 肽 FASTA → 对 Gmax/Gm2.1/GmZH/At marker 做 blastp 并合并结果 |
06_1.Individual_clusters_marker_gene_finder.sh |
取各簇 top-20 DEG 序列并抽 eggNOG 注释(启用簇 8) |
07_batch_multiple_scTenifoldKnk.sh |
每个候选基因生成一个 scTenifoldKnk 作业脚本与 scTenifoldKnk_run.list |
English
Purpose: Take the single-nucleus RNA-seq data on the Gifu T2T genome (the two root stages Dpi00 and Dpi02, two replicates each) from Cell Ranger quantification through clustering and cell-type annotation, cluster/cell-type DEGs and scTenifoldKnk virtual knockout.
Entry point: the seven shell scripts in the table below; the Script/*.py files with no entry script, run by hand, are 03_1-cluster_statitcs.py, 03_2-fliter_to_all_gene_h5ad.py, 05_3-marker_cluster_DEGs_own_order.py and makesure_02dpi_samples_all_have_Early_signal_cell.py.
Inputs: sample.list (ProjectID CNName RunID SampleName), the four *_filtered.h5ad objects, Gifu T2T FASTA/GTF and .pep.fa, {Gmax,Gm2.1,GmZH,At}_marker_genes.pep.fa, eggNOG annotation; later steps read 02_processData/ and 04_Anno/ from the working directory, and scTenifoldKnk additionally needs zz_00_pea.rds and All_species_merged_by_GifuT2T.SNF_Symbol.tsv.
Outputs: 00_ref/, 01_cellranger_out/<SampleName>/, 02_processData/ (h5ad files, zz_*, zz_00_pea.rds, 00_02_pseudobulk_fpkm.csv), 03_figures/ (dimension-reduction plots, clustree_plot_*.pdf, volcano plots), 04_Anno/ (Seurat rds, marker CSVs, blast_*/, enrichment results) and 06_batch_multiple_scTenifoldKnk/<group>/<gene>/.
Run: from the stage root (holding 02_processData/, 03_figures/, 04_Anno/) run the scripts in table order; the numbered scripts carry #CSUB directives and are submitted to the scheduler; for the {TOKEN} path placeholders see scripts/README.md; note that 03_1-dimenRedu_adjusted.py writes a file name that 03_1-cluster_statitcs.py and 03_2-fliter_to_all_gene_h5ad.py do not read, so it must be unified first.
Tools: Cell Ranger 9.0.1; Python 3.9.19 with scanpy/anndata/scvi-tools/leidenalg/rpy2; R 4.4.1 with Seurat/scTenifoldKnk/DESeq2/clustree; BLAST+ and seqkit; settings as coded: n_top_genes 4000 (scanned 1000–5000), Leiden 0.4 (scanned 0.2–1.2), and 100 scTenifoldKnk replicates per gene (K ∈ {10,30,50,70,100}).
| Script | Purpose |
|---|---|
00_mkref.sh |
cellranger mkref --genome=Gifu from the Gifu T2T FASTA + GTF → ./00_ref |
00_cellranger.sh |
one cellranger count per sample in sample.list → 01_cellranger_out/<SampleName> |
run_singlecell.sh |
one driver running Script/01_annaDataQC.py, 02_integrate.py, 03_1-dimenRedu_adjusted.py, 05_3-markerAnno_Ps_v5_own.py, 06_4-DEGs_cell_type.R in order |
05_cluster_DEGs_finder.sh |
05_1-…py then 05_2-…R: Seurat object and cluster markers, then per-cluster enrichment |
06_1.marker_gene_finder.sh |
DEGs → peptide FASTA → blastp against the Gmax/Gm2.1/GmZH/At markers, results merged |
06_1.Individual_clusters_marker_gene_finder.sh |
top-20 DEG sequences per cluster plus eggNOG annotation (active cluster 8) |
07_batch_multiple_scTenifoldKnk.sh |
one scTenifoldKnk job script per candidate gene plus scTenifoldKnk_run.list |
| 步骤 | 内容 |
|---|---|
| 单核分析脚本 | — |
以下为该步骤目录中的实际文件(同步快照,与该目录内容一致)。
00_cellranger.sh · Shell · 41 行
#!/bin/bash#CSUB -J cellranger#CSUB -q c01#CSUB -o cellranger.out#CSUB -e cellranger.error#CSUB -n 64#CSUB -R span[hosts=1]
REF={PROJ_LOTUS_HX}/07_singleCell/00_ref/fasta/genome.faDATADIR={HOME_HX}/Single_cell_dataSAMPLELIST=sample.list
# module load cellranger/7.2.0 # 如果集群环境要求,请加载软件
while read ProjectID CNName RunID SampleNamedo echo ">>> Processing $SampleName ($RunID)"
R1=${DATADIR}/${ProjectID}/${RunID}/${RunID}_S1_L001_R1_001.fastq.gz R2=${DATADIR}/${ProjectID}/${RunID}/${RunID}_S1_L001_R2_001.fastq.gz
if [[ ! -f $R1 ]]; then echo "❌ Missing R1 file: $R1" continue fi
outdir=./01_cellranger_out/${SampleName} mkdir -p ${outdir}
cellranger count \ --id=${SampleName} \ --transcriptome={PROJ_LOTUS_HX}/07_singleCell/00_ref \ --fastqs=$(dirname $R1) \ --sample=${RunID} \ --create-bam=true \ --localcores=32 \ --localmem=120 \ --chemistry=SC3Pv3 \ --output-dir=${outdir}
done < ${SAMPLELIST}00_mkref.sh · Shell · 17 行
#!/bin/bash#CSUB -J mkref#CSUB -q c01#CSUB -o mkref.out#CSUB -e mkref.error#CSUB -n 64#CSUB -R span[hosts=1]
reffasta={PROJ_LOTUS_HX}/00_assembly_result/GifuT2T/Lotus_GifuT2T_v1.0.fastarefgtf={PROJ_LOTUS_HX}/00_assembly_result/GifuT2T/Lotus_GifuT2T.gtf
cellranger mkref \ --genome=Gifu \ --fasta=$reffasta \ --genes=$refgtf \ --output-dir=./00_ref \ --nthreads=6405_cluster_DEGs_finder.sh · Shell · 76 行
#!/bin/bash#CSUB -J clusterDEG#CSUB -q c01#CSUB -o clusterDEG.out#CSUB -e clusterDEG.error#CSUB -n 64#CSUB -R span[hosts=1]
# 加载 condasource ~/anaconda3/etc/profile.d/conda.sh
# 激活环境conda activate singlecell
# 运行脚本python ./Script/05_1-Seurat_Exp_h5ad_to_rds.pyRscript ./Script/05_2-Exp_h5ad_to_rds_cluster.R
# 对每簇的marker genes的功能进行富集#========================# 0. 基本路径设置#========================output=./04_Anno/cluster_makers_genes_enrichmentmkdir -p "$output"
script={SOFTWARE_HX}/Script/RNA-seqanno={PROJ_LOTUS_HX}/00_assembly_result/GifuT2T/Lotus_Gifu.emapper.annotationsfile=./04_Anno/cluster_06_cluster_markers_gene1.csv
#========================# 1. 获取所有 cluster 编号#========================clusters=$(tail -n +2 "$file" \ | awk -F',' '{gsub(/"/,""); print $6}' \ | sort -u)
#========================# 2. 逐 cluster 筛选基因 + 富集#========================for cl in $clusters; do echo "处理 cluster ${cl}..."
# 固定文件名,不使用 mktemp,避免报错 tmpfile="${output}/cluster_${cl}.genelist"
# 筛选基因并替换 "-" 为 "_" awk -F',' -v c="$cl" ' BEGIN { OFS="\t" } NR>1 { gsub(/"/, "", $0) if ($6 == c && $2 > 1 && $5 < 0.05) { gsub("-", "_", $7) print $7 } } ' "$file" > "$tmpfile"
# 若无基因,则跳过 if [ ! -s "$tmpfile" ]; then echo " cluster ${cl} 没有满足条件的基因,跳过富集分析" rm "$tmpfile" continue fi
# cluster 单独输出目录 outdir="${output}/cluster_${cl}" mkdir -p "$outdir"
echo " 正在进行富集分析..." Rscript "${script}/Enrichment.R" "$anno" "$tmpfile" "$outdir"
done
echo "全部 cluster 处理完毕!"06_1.Individual_clusters_marker_gene_finder.sh · Shell · 55 行
#!/bin/bash#CSUB -J findmarker#CSUB -q c01#CSUB -o findmarker.out#CSUB -e findmarker.error#CSUB -n 64#CSUB -R span[hosts=1]
# 输入文件csv=./04_Anno/cluster_06_cluster_markers_gene.csvfasta={PROJ_LOTUS_HX}/00_assembly_result/GifuT2T/Lotus_GifuT2T.pep.faannot=./Lotus_Gifu.emapper.annotationsoutdir=./04_Anno
mkdir -p $outdir
# 需要处理的 cluster#clusters=("3" "4" "6")clusters=("8")
for n in "${clusters[@]}"; do echo ">>> Processing cluster $n ..."
# Step1: 获取 top10 基因列表 awk -F',' -v c="$n" ' NR>1 { gsub(/"/, "", $0); if ($6 == c) { gsub(/-/, "_", $7); print $2 "\t" $7 } }' "$csv" \ | sort -k1,1nr \ | head -n 20 \ | cut -f2 \ > ${outdir}/cluster_${n}_DEGs.list
# Step2: 从 pep.fa 提取序列 seqkit grep -f ${outdir}/cluster_${n}_DEGs.list "$fasta" \ > ${outdir}/cluster_${n}_DEGs.pep.fa
echo "Sequence extraction for cluster $n done."
# Step3: 从 EggNOG 注释中提取对应基因的注释 # 注释文件的第一列是 query ID grep -v "^#" "$annot" | awk -v OFS="\t" ' NR==FNR {a[$1]=1; next} ($1 in a) {print} ' ${outdir}/cluster_${n}_DEGs.list - \ > ${outdir}/cluster_${n}_DEGs.annotation
echo "Annotation extraction for cluster $n done."done
echo "All clusters finished."06_1.marker_gene_finder.sh · Shell · 175 行
#!/bin/bash#CSUB -J findmarker#CSUB -q c01#CSUB -o findmarker.out#CSUB -e findmarker.error#CSUB -n 64#CSUB -R span[hosts=1]
cul=`pwd`
# 输入文件csv=./04_Anno/cluster_06_cluster_markers_gene.csvfasta={PROJ_LOTUS_HX}/00_assembly_result/GifuT2T/Lotus_GifuT2T.pep.faout=./04_Anno/cluster_DEGs.pep.fa
# 对DEGs进行过滤,提取 gene 列,去掉表头并去重,并将-替换为_awk -F',' 'NR>1 { gsub(/"/, "", $0); if ( $5 < 0.05) { gsub(/-/, "_", $7); print $7 }}' "$csv" | sort -u > ./04_Anno/cluster_DEGs.list
# 用 seqkit 从 pep.fa 中提取对应序列seqkit grep -f ./04_Anno/cluster_DEGs.list "$fasta" > "$out"
echo "✅ Done! Extracted sequences saved to: $out"
outdir=./04_Anno/blast_Gmaxmkdir -p "$outdir"
# 构建blast数据库makeblastdb -in "./04_Anno/Gmax_marker_genes.pep.fa" -dbtype prot -out "$outdir/db"
# 运行blastpblastp -query "$out" -db "$outdir/db" \ -out "$outdir/result.blast" \ -evalue 1e-5 -outfmt "6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore qcovs" \ -num_threads 32
cut -f1,2,4,3 $outdir/result.blast > $outdir/result.blast.res
seqkit fx2tab -l -n -i ./04_Anno/Gmax_marker_genes.pep.fa > $outdir/ref_protein_lengths.txt
seqkit fx2tab -l -n -i $out > $outdir/query_protein_lengths.txt
cd $outdir
python3 $cul/blast_merge.py result.blast.res query_protein_lengths.txt ref_protein_lengths.txt blast1.merge
awk '$3>50 && $7>0.5 && $8>0.5' blast1.merge | \awk '{ diff = (1-$7>=0?1-$7:$7-1) + (1-$8>=0?1-$8:$8-1) key = $2 if (!(key in best) || $3 > best_id[key] || ($3 == best_id[key] && diff < best_diff[key])) { best[key] = $0 best_id[key] = $3 best_diff[key] = diff }}END { for (k in best) print best[k]}' > blast1.filtered.best
cd $culoutdir=./04_Anno/blast_Gm2.1mkdir -p "$outdir"
# 构建blast数据库makeblastdb -in "./04_Anno/Gm2.1_marker_genes.pep.fa" -dbtype prot -out "$outdir/db"
# 运行blastpblastp -query "$out" -db "$outdir/db" \ -out "$outdir/result.blast" \ -evalue 1e-5 -outfmt "6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore qcovs" \ -num_threads 32
cut -f1,2,4,3 $outdir/result.blast > $outdir/result.blast.res
seqkit fx2tab -l -n -i ./04_Anno/Gm2.1_marker_genes.pep.fa > $outdir/ref_protein_lengths.txt
seqkit fx2tab -l -n -i $out > $outdir/query_protein_lengths.txt
cd $outdir
python3 $cul/blast_merge.py result.blast.res query_protein_lengths.txt ref_protein_lengths.txt blast1.merge
awk '$3>50 && $7>0.5 && $8>0.5' blast1.merge | \awk '{ diff = (1-$7>=0?1-$7:$7-1) + (1-$8>=0?1-$8:$8-1) key = $2 if (!(key in best) || $3 > best_id[key] || ($3 == best_id[key] && diff < best_diff[key])) { best[key] = $0 best_id[key] = $3 best_diff[key] = diff }}END { for (k in best) print best[k]}' > blast1.filtered.best
cd $culoutdir=./04_Anno/blast_GmZHmkdir -p "$outdir"
# 构建blast数据库makeblastdb -in "./04_Anno/GmZH_marker_genes.pep.fa" -dbtype prot -out "$outdir/db"
# 运行blastpblastp -query "$out" -db "$outdir/db" \ -out "$outdir/result.blast" \ -evalue 1e-5 -outfmt "6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore qcovs" \ -num_threads 32
cut -f1,2,4,3 $outdir/result.blast > $outdir/result.blast.res
seqkit fx2tab -l -n -i ./04_Anno/GmZH_marker_genes.pep.fa > $outdir/ref_protein_lengths.txt
seqkit fx2tab -l -n -i $out > $outdir/query_protein_lengths.txt
cd $outdir
python3 $cul/blast_merge.py result.blast.res query_protein_lengths.txt ref_protein_lengths.txt blast1.merge
awk '$3>50 && $7>0.5 && $8>0.5' blast1.merge | \awk '{ diff = (1-$7>=0?1-$7:$7-1) + (1-$8>=0?1-$8:$8-1) key = $2 if (!(key in best) || $3 > best_id[key] || ($3 == best_id[key] && diff < best_diff[key])) { best[key] = $0 best_id[key] = $3 best_diff[key] = diff }}END { for (k in best) print best[k]}' > blast1.filtered.best
cd $culoutdir=./04_Anno/blast_Atmkdir -p "$outdir"
# 构建blast数据库makeblastdb -in "./04_Anno/At_marker_genes.pep.fa" -dbtype prot -out "$outdir/db"
# 运行blastpblastp -query "$out" -db "$outdir/db" \ -out "$outdir/result.blast" \ -evalue 1e-5 -outfmt "6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore qcovs" \ -num_threads 32
cut -f1,2,4,3 $outdir/result.blast > $outdir/result.blast.res
seqkit fx2tab -l -n -i ./04_Anno/At_marker_genes.pep.fa > $outdir/ref_protein_lengths.txt
seqkit fx2tab -l -n -i $out > $outdir/query_protein_lengths.txt
cd $outdir
python3 $cul/blast_merge.py result.blast.res query_protein_lengths.txt ref_protein_lengths.txt blast1.merge
awk '$3>50 && $7>0.5 && $8>0.5' blast1.merge | \awk '{ diff = (1-$7>=0?1-$7:$7-1) + (1-$8>=0?1-$8:$8-1) key = $2 if (!(key in best) || $3 > best_id[key] || ($3 == best_id[key] && diff < best_diff[key])) { best[key] = $0 best_id[key] = $3 best_diff[key] = diff }}END { for (k in best) print best[k]}' > blast1.filtered.best07_batch_multiple_scTenifoldKnk.sh · Shell · 59 行
#!/bin/bashset -euo pipefail
cul="$(pwd)"runlist="$cul/scTenifoldKnk_run.list"genelist=$1celltype=$2
# 清空 run list: > "$runlist"
while IFS=$'\t ' read -r gene group || [ -n "$gene" ]; do # 去掉 Windows CR gene="${gene//$'\r'/}" group="${group//$'\r'/}"
# 跳过非法行 [ -z "$gene" ] && continue [ -z "$group" ] && continue
# 构建目录:06_batch_multiple_scTenifoldKnk / group / gene base_dir="$cul/06_batch_multiple_scTenifoldKnk/$group/$gene" mkdir -p "$base_dir"
job="$base_dir/${gene}_scTenifoldKnk.sh"
cat > "$job" <<EOF#!/bin/bash#CSUB -J ${gene}_Knk#CSUB -q c02#CSUB -o $base_dir/${gene}_Knk.out#CSUB -e $base_dir/${gene}_Knk.error#CSUB -n 16#CSUB -R span[hosts=1]
set -eo pipefail # conda 前不要 -u
source ~/anaconda3/etc/profile.d/conda.shconda activate singlecell
set -euo pipefail # 激活后再启用 -u
# 进入 gene 专属目录cd "$base_dir"
# 软链接数据(避免重复拷贝)ln -sf $cul/02_processData/zz_00_pea.rdsln -sf $cul/All_species_merged_by_GifuT2T.SNF_Symbol.tsv
# 执行 scTenifoldKnkRscript $cul/Script/07_batch_multiple_scTenifoldKnk.R "$gene" "$celltype"EOF
chmod +x "$job" echo "$job" >> "$runlist"
done < "$genelist"
echo "Generated run list: $runlist"run_singlecell.sh · Shell · 43 行
#!/bin/bash#CSUB -J singlecell#CSUB -q c01#CSUB -o singlecell.out#CSUB -e singlecell.error#CSUB -n 64#CSUB -R span[hosts=1]
# Steps 01-03 and 06_2-06_3 of the single-nucleus pipeline, which were# previously one six-line wrapper each. The longer steps keep their own# scripts: 00_cellranger.sh, 00_mkref.sh, 05_cluster_DEGs_finder.sh,# 06_1.marker_gene_finder.sh, 06_1.Individual_clusters_marker_gene_finder.sh# and 07_batch_multiple_scTenifoldKnk.sh.## Run from 08_single_nucleus/:# bash run_singlecell.sh
# 加载 condasource ~/anaconda3/etc/profile.d/conda.sh
# 激活环境conda activate singlecell
# ── 01 QC ────────────────────────────────────────────────────────────────python ./Script/01_annaDataQC.py
# ── 02 integration ───────────────────────────────────────────────────────python ./Script/02_integrate.py
# ── 03 dimensionality reduction and clustering ───────────────────────────python ./Script/03_1-dimenRedu_adjusted.py
# ── 06_2 marker annotation ───────────────────────────────────────────────# 已有的文献的marker gene效果不好,故改用先发掘各簇DEG,在比对到大豆marker gene的思路。# python ./Script/05_3-marker_knowngene_lotus_own_order.pypython ./Script/05_3-markerAnno_Ps_v5_own.py
# ── 06_3 cell-type expression and DEGs ───────────────────────────────────# The three calls below feed this one and are commented out in the original run.# python ./Script/06_1-Seurat_Exp_h5ad_to_rds.py# Rscript ./Script/06_2-Exp_h5ad_to_rds_cell_type.R# Rscript ./Script/06_3-agg_to_fpkm.RRscript ./Script/06_4-DEGs_cell_type.R