See also
Paper describing expected error filtering and paired read merging (Edgar & Flyvbjerg, 2015).
Paired read assembler and quality filtering benchmark results
Read quality filtering
Expected errors
FASTQ format options
Quality scores
Choosing FASTQ filter parameters
Strategies for dealing with low-quality reverse reads (R2s)
The fastx_learn command is useful for checking the error rate after expected error quality filtering, which assumes that the Q scores are accurate. It does not use Q scores so gives an independent check.
Output options
| Option | Description | |
| -fastqout filename | FASTQ output file. You can use both -fastqout and -fastaout. |
|
| -fastaout filename | FASTA output file. You can use both -fastqout and -fastaout. |
|
| -fastqout_discarded filename | FASTQ output file for discarded reads. You can use both -fastqout_discarded and -fastaout_discarded. |
|
| -fastaout_discarded filename | FASTA output file for discarded reads. You can use both -fastqout_discarded and -fastaout_discarded. |
|
| -relabel prefix |
Generate new labels for the output sequences. They will be labeled prefix1, prefix2 and so on. For example, if you use -relabel SampleA. then the labels will be SampleA.1, SampleA.2 etc.
The special value @ indicates that the string should be constructed from the file name by truncating the file name at the first underscore or period and appending a period. With a typical Illumina FASTQ file name, this gives the sample name. So, for example, if the FASTQ file name is Mock_S188_L001_R1_001.fastq, then the string is Mock and the output labels will be Mock.1, Mock.2 etc. |
|
| -fastq_eeout | Append the expected number of errors according to the Q scores to the label in the format "ee=xx;". Expected errors are calculated after truncation, if applicable. |
|
| -sample string | Append sample=string; to the read label. |
Filtering options
| Option | Description | |
| -fastq_truncqual N | Truncate the read at the first position having quality score <= N, so that all remaining Q scores are >N. |
|
| -fastq_maxee E | Discard reads with > E total expected errors for all bases in the read after any truncation options have been applied. |
|
| -fastq_trunclen L | Truncate sequences at the L'th base. If the sequence is shorter than L, discard. |
|
| -fastq_minlen L | Discard sequences with < L letters. |
|
| -fastq_stripleft N | Delete the first N bases in the read. |
|
| -fastq_maxee_rate E | Discard reads with > E expected errors per base.Calculated after any truncation options have been applied. For example, with the fastq_maxee_rate option set to 0.01, then a read of length 100 will be discarded if the expected errors is >1, and a read of length 1,000 will be discarded if the expected errors is >10. |
|
| -fastq_maxns k | Discard if there are >k Ns in the read. |
Examples
"Raw" conversion of Sanger FASTQ to FASTA with no filtering:
usearch -fastq_filter reads.fastq -fastaout reads.fasta -fastq_ascii 64
Truncate to length 150, discard if expected errors > 0.5, and convert to FASTA:
usearch -fastq_filter reads.fastq -fastq_trunclen 150 -fastq_maxee 0.5 \
-fastaout reads.fasta
Truncate a read at length 100 and then discard if it contains a Q<15, output to new FASTQ file:
usearch -fastq_filter reads.fastq -fastq_minlen 100 -fastq_truncqual 15 \