安装Rust/Cargo
curl --proto '=https' --tlsv1.2 -sSf <https://sh.rustup.rs> | sh
或者升级，推荐升级到1.77.0
rustup update stable
已编译可多平台/跨平台使用
因为用到了SIMD加速，所以在不同CPU上会有差异：
在Intel/AMD x86 架构CPU上
修改Cargo.toml
[package.metadata.docs.rs]
features = ["simd_avx2"]
[target.'cfg(target_arch = "aarch64")'.dependencies]
block-aligner = { version = "0.5", features = ["simd_avx2"] }
编译：
cargo build --features simd_avx2 --release
理论上只要内存吃得消读入FASTQ或线程数 * batch_reads所需要的内存，不必担心内存溢出的问题，因此可以尽可能大量使用线程数处理。
segpe --help
Algorithmic ideas of SegPE:
- Exact match search: This is a more direct method for finding exact matches of artificial sequences.
Regular expression matching and Hamming distance:
This is suitable for detecting and locating index sequences, especially when mismatches of a certain length are taken into account.
- Process and classify PE sequences: After removing the adapter and index sequences, classify the PE sequences and create new PE FASTQ files.
- Use Multi-threads, SIMD and AsyncIO to handle large amounts of data.
Usage: segpe [OPTIONS] --five-art-fa <FIVE_ART_FA> --three-art-fa <THREE_ART_FA> --five-idx-fa <FIVE_IDX_FA> --pe1-fastq <PE1_FASTQ> --pe2-fastq <PE2_FASTQ>
Options:
--five-art-fa <FIVE_ART_FA>
Path of 5' artificial fasta file
--three-art-fa <THREE_ART_FA>
Path of 3' artificial fasta file
--five-idx-fa <FIVE_IDX_FA>
Path of 5‘ index fasta file
--three-idx-fa <THREE_IDX_FA>
Path of 3’ index fasta file
[default: ]
--idx-loc <IDX_LOC>
Location of index, 1: PE1 2: PE2 3: both
[default: 1]
--pe1-fastq <PE1_FASTQ>
Path of PE1 fastq file
--pe2-fastq <PE2_FASTQ>
Path of PE1 fastq file
-s, --seed-len <SEED_LEN>
Number of seed length, not allow to longer than index length
[default: 6]
--merge-pe
Merged overlapped PE
--error-tolerance <ERROR_TOLERANCE>
Number of error tolerance
[default: 1]
-m, --match-score <MATCH_SCORE>
Number of alignment macth score
[default: 1]
--error-score <ERROR_SCORE>
Number of alignment mismacth score
[default: -1]
--gap-open-score <GAP_OPEN_SCORE>
Number of alignment gap open score
[default: -5]
--gap-extend-score <GAP_EXTEND_SCORE>
Number of alignment gap extend score
[default: -1]
-n, --num-threads <NUM_THREADS>
Number of cucurrency threads
[default: 8]
-b, --batch <BATCH>
batch size of reads, which every thread need to handle
[default: 10000]
--train <TRAIN>
pretrain size of reads
[default: 0]
--trim-name
add trim info in reads_name
-o, --outdir <OUTDIR>
Path of output directory
[default: ./]
-h, --help
Print help (see a summary with '-h')
-V, --version
Print version
https://excalidraw.com/#room=735721d320c8e703d672,MonxP9T7XWpM2JvdqAiTdg
| 参数 | 说明 |
|---|---|
| --five-art-fa <FIVE_ART_FA> | 输入建库片段5‘段接的人工序列，包括：Sequencing Adapter/Primer, Barcode, Index, Key Sequence，按接入序列从5‘->3'的测序方向连接生成的fasta序列，包括PE1/PE2 5’端会测到的人工序列 |
| --three-art-fa | 输入建库片段3‘段接的人工序列，包括：Sequencing Adapter/Primer, Barcode, Index, Key Sequence，按接入序列从5‘->3'的测序方向连接生成的fasta序列, 包括PE1/PE2 3’端(包括merge PE后的3‘端)会测到的人工序列 |
| --five-idx-fa <FIVE_IDX_FA> | 放在5‘端的索引序列，用于分库的序列，不管是PE1还是PE2的5’端，同时这个序列应该出现在five-art-fa |
| --three-idx-fa <THREE_IDX_FA> | 放在3‘端的索引序列，用于分库的序列，不管是PE1还是PE2的3’端同时这个序列应该出现在three-art-fa |
| --idx-loc <IDX_LOC> | 指定用于判断是PE1的索引序列，还是PE2的索引序列，还是同时一起判断。1: PE1 2: PE2 3: both默认值：1，即只用PE1的索引序列做判定 |
| --pe1-fastq <PE1_FASTQ> | 输入PE1的序列FASTQ文件 |
| --pe2-fastq <PE2_FASTQ> | 输入PE2的序列FASTQ文件 |
| -s, --seed-len <SEED_LEN> | 用于检索的最小seed值默认值：6 |
| --merge-pe <MERGE_PE> | 设置是否合并测穿有overlapping的PE序列默认值：false |
| --error-tolerance <ERROR_TOLERANCE> | 比对大容错值默认值：1 |
| -m, --match-score <MATCH_SCORE> | 比对打分矩阵中匹配碱基的分值默认值：1 |
| --error-score <ERROR_SCORE> | 比对打分矩阵中错配碱基的分值默认值：-1 |
| --gap-open-score <GAP_OPEN_SCORE> | 比对打分矩阵中打开gap的分值默认值：-5 |
| --gap-extend-score <GAP_EXTEND_SCORE> | 比对打分矩阵中gap延伸的分值默认值：-1 |
| -n, --num-threads <NUM_THREADS> | 并行线程数默认值: 8 |
| -b, --batch <BATCH> | 一个线程处理的reads数默认值：10000 |
| --train <TRAIN> | 预训练数据，用前多个reads做预训练，将比对分选的art序列存储起来，在实际跑数据时，如果匹配，就无须再进行比对分数据了。默认值：0 （即不进行预训练） |
| --trim-name | 当需要将trim掉的人工序列的位置信息输出到序列名中，选择此参数，默认不再修改序列名(v0.1.13版本以上） |
| -o, --outdir <OUTDIR> | 输出路径默认值：./⚠️注意：如果输出路径有已命名的相同输出文件的话，会追加生成。 |