2025/09/19 追記
fastpのversion 1.0がリリースされ、フォルダ内のfastqをバッチ処理する便利なスクリプトも提供されました。このスクリプトの使い方を確認しておきます。
インストール
最新のfastpにパスが通っている必要がある。fastpのバージョンが1未満だと動作しない。
#bioconda (link) v.1系指定、最新バージョンはリンク先確認
mamba install -c bioconda -y fastp==1.0.1
# download the latest build
wget http://opengene.org/fastp/fastp
chmod a+x ./fastp
#parallel.pyはcondaではインストールされない。レポジトリから取得する。
git clone https://github.com/OpenGene/fastp.git
#パスの通ったディレクトリに移動させる
chmod +x fastp/parallel.py
mv fastp/parallel.py /usr/local/bin/ #(あるいは~/binなど)
> python parallel.py -h
Usage: A python script to use fastp to preprocess all FASTQ files within a folder
Options:
--version show program's version number and exit
-h, --help show this help message and exit
-i INPUT_DIR, --input_dir=INPUT_DIR
the folder contains the FASTQ files to be
preprocessed, by default is current dir (.)
-o OUT_DIR, --out_dir=OUT_DIR
the folder to store the clean FASTQ. If not specified,
then there will be no output files.
-r REPORT_DIR, --report_dir=REPORT_DIR
the folder to store QC reports. If not specified, use
out_dir if out_dir is specified, otherwise use
input_dir.
-c COMMAND, --command=COMMAND
the path to fastp command, if not specified, then it
will use 'fastp' in PATH
-a ARGS, --args=ARGS the arguments that will be passed to fastp. Enclose in
quotation marks. Like --args='-f 3 -t 3'
-p PARALLEL, --parallel=PARALLEL
the number of fastp processes can be run in parallel,
if not specified, then it will be CPU_Core/4
-1 READ1_FLAG, --read1_flag=READ1_FLAG
specify the name flag of read1, default is R1, which
means a file with name *R1* is read1 file
-2 READ2_FLAG, --read2_flag=READ2_FLAG
specify the name flag of read2, default is R2, which
means a file with name *R2* is read2 file
実行方法
fastq(fastq.gz)を含むディレクトリ、出力ディレクトリ、HTMLとJSON形式レポートの保存ディレクトリをそれぞれ指定する。fastpのネイティブオプションのコマンドは-a ' 'で指定する。例えば-f3 -t2は各fastqの先頭3bpと末尾2bpをそれぞれ強制トリミングする。
python parallel.py -i input_fastq_dir -o fq_outdir -r report_outdir -a '-f 3 -t 2'
- -i the folder contains the FASTQ files to be preprocessed, by default is current dir (.)
- -o the folder to store the clean FASTQ. If not specified, then there will be no output files.
- -r the folder to store QC reports. If not specified, use out_dir if out_dir is specified, otherwise use input_dir.
- -a the arguments that will be passed to fastp. Enclose in quotation marks. Like --args='-f 3 -t 3'
- -p the number of fastp processes can be run in parallel, if not specified, then it will be CPU_Core/4
- -1 specify the name flag of read1, default is R1, which means a file with name *R1* is read1 file
- -2 specify the name flag of read2, default is R2, which means a file with name *R2* is read2 file
結果は指定したフォルダ内に保存される。ペアエンドのリードはデフォルトではR1とR2が認識される。変更するには-1と-2オプションを使う。
parallel.py \
-i ./ -o fastp_processed -r report -p 6 \
-1 "_R1" -2 "_R2" \
-a "-q 20 -u 30 -n 10 -l 20 --correction -w 3"
-
-c, --correction enable base correction in overlapped regions (only for PE data), default is disabled
-
-u, --unqualified_percent_limit how many percents of bases are allowed to be unqualified (0~100). Default 40 means 40% (int [=40])
その他
- ペアエンドの片方しか認識されなかったりルールとファイル名が異なる場合、スキップされるわけではなく2つのシングルエンドとして処理される(同期が取れなくなっている可能性が高いので注意)。
- この挙動から、シングルエンドとペアエンドが混在していても全てのfastqをバッチ処理できると考えられる。
引用
fastp 1.0: An ultra-fast all-round tool for FASTQ data quality control and preprocessing
Shifu Chen
iMeta, First published: 09 September 2025.
関連
v1.0のoptipn
$ fastp
fastp: an ultra-fast all-in-one FASTQ preprocessor
version 1.0.1
usage: fastp [options] ...
options:
-i, --in1 read1 input file name (string [=])
-o, --out1 read1 output file name (string [=])
-I, --in2 read2 input file name (string [=])
-O, --out2 read2 output file name (string [=])
--unpaired1 for PE input, if read1 passed QC but read2 not, it will be written to unpaired1. Default is to discard it. (string [=])
--unpaired2 for PE input, if read2 passed QC but read1 not, it will be written to unpaired2. If --unpaired2 is same as --unpaired1 (default mode), both unpaired reads will be written to this same file. (string [=])
--overlapped_out for each read pair, output the overlapped region if it has no any mismatched base. (string [=])
--failed_out specify the file to store reads that cannot pass the filters. (string [=])
-m, --merge for paired-end input, merge each pair of reads into a single read if they are overlapped. The merged reads will be written to the file given by --merged_out, the unmerged reads will be written to the files specified by --out1 and --out2. The merging mode is disabled by default.
--merged_out in the merging mode, specify the file name to store merged output, or specify --stdout to stream the merged output (string [=])
--include_unmerged in the merging mode, write the unmerged or unpaired reads to the file specified by --merge. Disabled by default.
-6, --phred64 indicate the input is using phred64 scoring (it'll be converted to phred33, so the output will still be phred33)
-z, --compression compression level for gzip output (1 ~ 9). 1 is fastest, 9 is smallest, default is 4. (int [=4])
--stdin input from STDIN. If the STDIN is interleaved paired-end FASTQ, please also add --interleaved_in.
--stdout stream passing-filters reads to STDOUT. This option will result in interleaved FASTQ output for paired-end output. Disabled by default.
--interleaved_in indicate that <in1> is an interleaved FASTQ which contains both read1 and read2. Disabled by default.
--reads_to_process specify how many reads/pairs to be processed. Default 0 means process all reads. (int [=0])
--dont_overwrite don't overwrite existing files. Overwritting is allowed by default.
--fix_mgi_id the MGI FASTQ ID format is not compatible with many BAM operation tools, enable this option to fix it.
-V, --verbose output verbose log information (i.e. when every 1M reads are processed).
-A, --disable_adapter_trimming adapter trimming is enabled by default. If this option is specified, adapter trimming is disabled
-a, --adapter_sequence the adapter for read1. For SE data, if not specified, the adapter will be auto-detected. For PE data, this is used if R1/R2 are found not overlapped. (string [=auto])
--adapter_sequence_r2 the adapter for read2 (PE data only). This is used if R1/R2 are found not overlapped. If not specified, it will be the same as <adapter_sequence> (string [=auto])
--adapter_fasta specify a FASTA file to trim both read1 and read2 (if PE) by all the sequences in this FASTA file (string [=])
-2, --detect_adapter_for_pe enable adapter detection for PE data to get ultra-clean data. It takes more time to find just a little bit more adapters.
--allow_gap_overlap_trimming allow up to one gap when trim adapters by overlap analysis for PE data. By default no gap is allowed.
-f, --trim_front1 trimming how many bases in front for read1, default is 0 (int [=0])
-t, --trim_tail1 trimming how many bases in tail for read1, default is 0 (int [=0])
-b, --max_len1 if read1 is longer than max_len1, then trim read1 at its tail to make it as long as max_len1. Default 0 means no limitation (int [=0])
-F, --trim_front2 trimming how many bases in front for read2. If it's not specified, it will follow read1's settings (int [=0])
-T, --trim_tail2 trimming how many bases in tail for read2. If it's not specified, it will follow read1's settings (int [=0])
-B, --max_len2 if read2 is longer than max_len2, then trim read2 at its tail to make it as long as max_len2. Default 0 means no limitation. If it's not specified, it will follow read1's settings (int [=0])
-D, --dedup enable deduplication to drop the duplicated reads/pairs
--dup_calc_accuracy accuracy level to calculate duplication (1~6), higher level uses more memory (1G, 2G, 4G, 8G, 16G, 24G). Default 1 for no-dedup mode, and 3 for dedup mode. (int [=0])
--dont_eval_duplication don't evaluate duplication rate to save time and use less memory.
-g, --trim_poly_g force polyG tail trimming, by default trimming is automatically enabled for Illumina NextSeq/NovaSeq data
--poly_g_min_len the minimum length to detect polyG in the read tail. 10 by default. (int [=10])
-G, --disable_trim_poly_g disable polyG tail trimming, by default trimming is automatically enabled for Illumina NextSeq/NovaSeq data
-x, --trim_poly_x enable polyX trimming in 3' ends.
--poly_x_min_len the minimum length to detect polyX in the read tail. 10 by default. (int [=10])
-5, --cut_front move a sliding window from front (5') to tail, drop the bases in the window if its mean quality < threshold, stop otherwise.
-3, --cut_tail move a sliding window from tail (3') to front, drop the bases in the window if its mean quality < threshold, stop otherwise.
-r, --cut_right move a sliding window from front to tail, if meet one window with mean quality < threshold, drop the bases in the window and the right part, and then stop.
-W, --cut_window_size the window size option shared by cut_front, cut_tail or cut_sliding. Range: 1~1000, default: 4 (int [=4])
-M, --cut_mean_quality the mean quality requirement option shared by cut_front, cut_tail or cut_sliding. Range: 1~36 default: 20 (Q20) (int [=20])
--cut_front_window_size the window size option of cut_front, default to cut_window_size if not specified (int [=4])
--cut_front_mean_quality the mean quality requirement option for cut_front, default to cut_mean_quality if not specified (int [=20])
--cut_tail_window_size the window size option of cut_tail, default to cut_window_size if not specified (int [=4])
--cut_tail_mean_quality the mean quality requirement option for cut_tail, default to cut_mean_quality if not specified (int [=20])
--cut_right_window_size the window size option of cut_right, default to cut_window_size if not specified (int [=4])
--cut_right_mean_quality the mean quality requirement option for cut_right, default to cut_mean_quality if not specified (int [=20])
-Q, --disable_quality_filtering quality filtering is enabled by default. If this option is specified, quality filtering is disabled
-q, --qualified_quality_phred the quality value that a base is qualified. Default 15 means phred quality >=Q15 is qualified. (int [=15])
-u, --unqualified_percent_limit how many percents of bases are allowed to be unqualified (0~100). Default 40 means 40% (int [=40])
-n, --n_base_limit if one read's number of N base is >n_base_limit, then this read/pair is discarded. Default is 5 (int [=5])
-e, --average_qual if one read's average quality score <avg_qual, then this read/pair is discarded. Default 0 means no requirement (int [=0])
-L, --disable_length_filtering length filtering is enabled by default. If this option is specified, length filtering is disabled
-l, --length_required reads shorter than length_required will be discarded, default is 15. (int [=15])
--length_limit reads longer than length_limit will be discarded, default 0 means no limitation. (int [=0])
-y, --low_complexity_filter enable low complexity filter. The complexity is defined as the percentage of base that is different from its next base (base[i] != base[i+1]).
-Y, --complexity_threshold the threshold for low complexity filter (0~100). Default is 30, which means 30% complexity is required. (int [=30])
--filter_by_index1 specify a file contains a list of barcodes of index1 to be filtered out, one barcode per line (string [=])
--filter_by_index2 specify a file contains a list of barcodes of index2 to be filtered out, one barcode per line (string [=])
--filter_by_index_threshold the allowed difference of index barcode for index filtering, default 0 means completely identical. (int [=0])
-c, --correction enable base correction in overlapped regions (only for PE data), default is disabled
--overlap_len_require the minimum length to detect overlapped region of PE reads. This will affect overlap analysis based PE merge, adapter trimming and correction. 30 by default. (int [=30])
--overlap_diff_limit the maximum number of mismatched bases to detect overlapped region of PE reads. This will affect overlap analysis based PE merge, adapter trimming and correction. 5 by default. (int [=5])
--overlap_diff_percent_limit the maximum percentage of mismatched bases to detect overlapped region of PE reads. This will affect overlap analysis based PE merge, adapter trimming and correction. Default 20 means 20%. (int [=20])
-U, --umi enable unique molecular identifier (UMI) preprocessing
--umi_loc specify the location of UMI, can be (index1/index2/read1/read2/per_index/per_read, default is none (string [=])
--umi_len if the UMI is in read1/read2, its length should be provided (int [=0])
--umi_prefix if specified, an underline will be used to connect prefix and UMI (i.e. prefix=UMI, UMI=AATTCG, final=UMI_AATTCG). No prefix by default (string [=])
--umi_skip if the UMI is in read1/read2, fastp can skip several bases following UMI, default is 0 (int [=0])
--umi_delim delimiter to use between the read name and the UMI, default is : (string [=:])
-p, --overrepresentation_analysis enable overrepresented sequence analysis.
-P, --overrepresentation_sampling one in (--overrepresentation_sampling) reads will be computed for overrepresentation analysis (1~10000), smaller is slower, default is 20. (int [=20])
-j, --json the json format report file name (string [=fastp.json])
-h, --html the html format report file name (string [=fastp.html])
-R, --report_title should be quoted with ' or ", default is "fastp report" (string [=fastp report])
-w, --thread worker thread number, default is 3 (int [=3])
-s, --split split output by limiting total split file number with this option (2~999), a sequential number prefix will be added to output name ( 0001.out.fq, 0002.out.fq...), disabled by default (int [=0])
-S, --split_by_lines split output by limiting lines of each file with this option(>=1000), a sequential number prefix will be added to output name ( 0001.out.fq, 0002.out.fq...), disabled by default (long [=0])
-d, --split_prefix_digits the digits for the sequential number padding (1~10), default is 4, so the filename will be padded as 0001.xxx, 0 to disable padding (int [=4])
--cut_by_quality5 DEPRECATED, use --cut_front instead.
--cut_by_quality3 DEPRECATED, use --cut_tail instead.
--cut_by_quality_aggressive DEPRECATED, use --cut_right instead.
--discard_unmerged DEPRECATED, no effect now, see the introduction for merging.
-?, --help print this message
Citation:
Shifu Chen. 2023. Ultrafast one-pass FASTQ data preprocessing, quality control, and deduplication using fastp. iMeta 2: e107