PERF is a Python package developed for fast and accurate identification of microsatellites from DNA sequences. Microsatellites or Simple Sequence Repeats (SSRs) are short tandem repeats of 1-6nt motifs. They are present in all genomes, and have a wide range of uses and functional roles. The existing tools for SSR identification have one or more caveats in terms of speed, comprehensiveness, accuracy, ease-of-use, flexibility and memory usage. PERF was designed to address all these problems.
PERF is a recursive acronym that stands for "PERF is an Exhaustive Repeat Finder". It is compatible with both Python 2 (tested on Python 2.7) and 3 (tested on Python 3.5). Its key features are:
- Fast run time. As an example, identification of all SSRs from the entire human genome takes less than 7 minutes. The speed can be further improved ~3 to 4 fold using PyPy (human genome finishes in less than 2 minutes using PyPy v5.8.0)
- Linear time and space complexity (O(n))
- Identifies perfect SSRs
- 100% accurate and comprehensive - Does not miss any repeats or does not pick any incorrect ones
- Easy to use - The only required argument is the input DNA sequence in FASTA format
- Flexible - Most of the parameters are customizable by the user at runtime
- Repeat cutoffs can be specified either in terms of the total repeat length or in terms of number of repeating units
- TSV output and HTML report. The default output is an easily parseable and exportable tab-separated format. Optionally, PERF also generates an interactive HTML report that depicts trends in repeat data as concise charts and tables
- Fixed usage of unit options file input for fastq input.
- Fixed usage of repeats input file.
- Annotation of repeats w.r.t to genomic context using a GFF or GTF file. (option -g).
- Multi-threading. Parallel identification of repeats in different sequences.
- Identification of perfect repeats in fastq files.
- Analysis report for repeats in fastq files.
- Option to identify atomic repeats.
- Analysis report rebuilt with Semantic ui and Apex Charts.
- Visualisation of repeat annotation data in analysis report.
- Python2 compatability fixed.
- Bug fixes for PyPi compatability.
- Import error issues.
PERF can be directly installed using pip with the package name perf_ssr
.
$ pip install perf_ssr
This name was chosen for the package so as not to clash with the existing perf
package.
Alternatively, it can be installed from the source code:
# Download the git repo
$ git clone https://github.com/RKMlab/perf.git
# Install
$ cd perf
$ python setup.py install
Both of the methods add a console command PERF
, which can be executed from any directory. It can also be used without installation by running the core.py
file in the PERF
subfolder:
$ git clone https://github.com/RKMlab/perf.git
$ cd perf/PERF
$ python core.py -h # Print the help message of PERF (see below)
The help message and available options can be accessed using
$ PERF -h # Short option
$ PERF --help # Long option
which gives the following output
usage: core.py [-h] -i <FILE> [-o <FILE>] [--format <STR>] [--version]
[-rep <FILE>] [-m <INT>] [-M <INT>] [-s <INT>] [-S <FLOAT>]
[--include-atomic] [-l <INT> | -u INT or FILE] [-a] [--info]
[-g <FILE>] [--anno-format <STR>] [--gene-key <STR>]
[--up-promoter <INT>] [--down-promoter <INT>]
[-f <FILE> | -F <FILE>] [-t <INT>]
Required arguments:
-i <FILE>, --input <FILE>
Input sequence file.
Optional arguments:
-o <FILE>, --output <FILE>
Output file name. Default: Input file name + _perf.tsv
--format <STR> Input file format. Default: fasta, Permissible: fasta,
fastq
--version show program's version number and exit
-rep <FILE>, --repeats <FILE>
File with list of repeats (Not allowed with -m and/or
-M)
-m <INT>, --min-motif-size <INT>
Minimum size of a repeat motif in bp (Not allowed with
-rep)
-M <INT>, --max-motif-size <INT>
Maximum size of a repeat motif in bp (Not allowed with
-rep)
-s <INT>, --min-seq-length <INT>
Minimum size of sequence length for consideration (in
bp)
-S <FLOAT>, --max-seq-length <FLOAT>
Maximum size of sequence length for consideration (in
bp)
--include-atomic An option to include factor atomic repeats for minimum
motif sizes greater than 1.
-l <INT>, --min-length <INT>
Minimum length cutoff of repeat
-u INT or FILE, --min-units INT or FILE
Minimum number of repeating units to be considered.
Can be an integer or a file specifying cutoffs for
different motif sizes.
-a, --analyse Generate a summary HTML report.
--info Sequence file info recorded in the output.
-f <FILE>, --filter-seq-ids <FILE>
List of sequence ids in fasta file which will be
ignored.
-F <FILE>, --target-seq-ids <FILE>
List of sequence ids in fasta file which will be used.
-t <INT>, --threads <INT>
Number of threads to run the process on. Default is 1.
Annotation arguments:
-g <FILE>, --annotate <FILE>
Genic annotation input file for annotation, Both GFF
and GTF can be processed. Use --anno-format to specify
format.
--anno-format <STR> Format of genic annotation file. Valid inputs: GFF,
GTF. Default: GFF
--gene-key <STR> Attribute key for geneId. The default identifier is
"gene". Please check the annotation file and pick a
robust gene identifier from the attribute column.
--up-promoter <INT> Upstream distance(bp) from TSS to be considered as
promoter region. Default 1000
--down-promoter <INT>
Downstream distance(bp) from TSS to be considered as
promoter region. Default 1000
The details of each option are given below:
Expects: FILE
Default: None
This is the only required argument for the program. The input file must be a valid FASTA/FASTQ file. PERF uses Biopython's FASTA parser to read the input fasta files. It accepts both single-line and multi-line sequences. Files with multiple sequences are also valid. To see more details about the FASTA format, see this page.
Expects: STRING (to be used as filename)
Default: Input Filename + _perf.tsv (see below)
If this option is not provided, the default output filename will be the same as the input filename, with its extension replaced with '_perf.tsv'. For example, if the input filename is my_seq.fa
, the default output filename will be my_seq_perf.tsv
. If the input filename does not have any extension, _perf.tsv
will be appended to the filename. Please note that even in the case of no identified SSRs, the output file is still created (therefore overwriting any previous file of the same name) but with no content in the file.
The output is a tab-delimited file, with one SSR record per line. The output columns follow the BED format. The details of the columns are given below:
S.No | Column | Description |
---|---|---|
1 | Chromosome | Chromosome or Sequence Name as specified by the first word in the FASTA header |
2 | Repeat Start | 0-based start position of SSR in the Chromosome |
3 | Repeat Stop | End position of SSR in the Chromosome |
4 | Repeat Class | Class of repeat as grouped by their cyclical variations |
5 | Repeat Length | Total length of identified repeat in nt |
6 | Repeat Strand | Strand of SSR based on their cyclical variation |
7 | Motif Number | Number of times the base motif is repeated |
8 | Actual Repeat | Starting sequence of the SSR irrespective of Repeat class and strand |
An example output showing some of the largest repeats from Drosophila melanogaster is given below
X 22012826 22014795 ACTGGG 1969 - 328 TCCCAG
2RHet 591337 591966 AATACT 629 - 104 ATTAGT
4 1042143 1042690 AAATAT 547 + 91 AAATAT
2RHet 598244 598789 AATACT 545 - 90 AGTATT
XHet 122 663 AGAT 541 + 135 GATA
X 22422335 22422827 AGAT 492 + 123 GATA
3R 975265 975710 AAAT 445 - 111 TTAT
X 15442288 15442724 ACAGAT 436 + 72 ACAGAT
2L 22086818 22087152 AATACT 334 - 55 TATTAG
YHet 137144 137466 AAGAC 322 - 64 CTTGT
The output is a tab-delimited file, with data on each repeat class per line.
S.No | Column | Description |
---|---|---|
1 | Repeat Class | Class of repeat as grouped by their cyclical variations |
2 | Number of reads | Number of reads having an instance of the repeat |
3 | Frequency | Total number of instances of the repeat |
4 | Bases | Total number of bases covered by the repeat |
5 | Repeat reads per million reads | Number of |
6 | Instances per million reads | Strand of SSR based on their cyclical variation |
7 | Repeat Bases per MB of sequence | Number of times the base motif is repeated |
8 | Length distribution | Starting sequence of the SSR irrespective of Repeat class and strand |
9 | Motif distribution | Starting sequence of the SSR irrespective of Repeat class and strand |
Expects: STRING (specifying format of the file)
Default: fasta
PERF was originally developed to identify repeats in FASTA files. In version 4.0.0 PERF can identify repeats in FASTQ sequence files as well. The default format the program expects is fasta. Specify input format as 'fasta' for FASTA files and 'fastq' for FASTQ files.
Expects: None
Default: False
In addition to the default tab-separated output, PERF can also generate a fully interactive HTML report for easy downstream analysis of the repeat data. The filename will be the same prefix as that of the main output. For example, if the input filename was my_seq.fa
, the analysis report will be my_seq_perf.html
. An example HTML report, generated from the repeat data of Homo sapiens (build hg19), can be accessed here (Right click -> Save As).
Expects: INTEGER
Default: 12
Minimum length cut-off to be considered when finding an SSR. The same cut-off will apply for SSRs of all motif lengths, even if the motif length is not a divisor of this value. In such cases, SSRs that end with a partial motif are also picked if they pass the length cut-off.
Expects: INTEGER OR FILE
Default: None
This option finds SSRs with a minimum number of repeating motifs. The argument accepts either an integer or file. If an integer is specified, the same value is used for all motif lengths. Instead, a specific value for each motif length using a two-column tab-separated file as demonstrated below:
$ cat repeat_units.txt
1 10
2 6
3 4
4 3
5 2
6 2
The first column specifies the motif length, and the second column specifies the minimum number of times the motif should be repeated to be considered an SSR. This file can be used to identify repeats with different number of repeating motifs: monomers repeated at least 10 times, dimers repeated at least 6 times etc., using the following command
$ PERF -i my_seq.fa -m 1 -M 6 -u repeat_units.txt
Expects: FILE
Default: None
PERF provides an option to limit the search to specific repeat motifs. The repeats of interest should be specified via a file containing 4 tab-separated columns, as shown below:
$ cat my_repeats.txt
A A 1 +
T A 1 -
AG AG 2 +
CT AG 2 -
GA AG 2 +
TC AG 2 -
$ PERF -i my_seq.fa -rep my_repeats.txt # Find all A and AG repeats from my_seq.fa
Note: This option is not allowed when -m
or -M
options are used.
Expects: INTEGER
Default: 1
Minimum length of motifs to be considered. By default, PERF ignores redundant motifs. For example, a stretch of 12 A's is considered a monomer repeat of 12 A's rather than a dimer repeat of 6 AA's. However, this is only true if -m
is set to 1. If for example, -m
is set to 2, then stretches of 12 A's are reported as dimer AA repeats. If this behavior isn't desired, we suggest using the -rep
option (see above) to specify the motifs that should/shouldn't be included.
Expects: INTEGER
Default: 6
Maximum length of motifs to be considered. Setting a large value of -M
has a non-trivial effect on both the runtime and memory usage of PERF. This is noticeable with -M
values above 10.
Expects: None
Default: False
Searches for atomic repeats when set to True. For example, when minimum motif size is set to 2bp, PERF ignores monomer repeats. When include atomic repeats is set to True, PERF identifies AA, CC, GG and TT as dimer repeats.
Expects: INTEGER
Default: 0
Minimum length of the input sequence to be searched for SSRs in bp. All sequences in the input file that are smaller than this length will be ignored.
Expects: INTEGER
Default: Infinity
Maximum length of the input sequence to be searched for SSRs in bp. All sequences in the input file that are larger than this length will be ignored.
Expects: FILE
Default: None
This option accepts a file with a list of sequence IDs in the input file that should be ignored. Useful for ignoring contigs, scaffold, or other poor quality sequences. The IDs can be FASTA headers (starting with '>' symbol) or just the names without the '>' symbol.
Expects: FILE
Default: None
This option accepts a file with a list of sequence IDs in the input file that should be analyzed. All other sequences will be ignored. Useful for analyzing specific chromosomes from a large input file. The IDs can be FASTA headers (starting with '>' symbol) or just the names without the '>' symbol.
Expects: None
Default: False
This option when set to True, includes information about the input sequence files and repeat summary data in the output file.
$ tail -5 test_input_perf.tsv
gi|514486271|gb|KE346361.1| 2667759 2667775 ATC 16 + 5 CAT
#File_name: test_input.fa
#Total_sequences: 2
#Total_bases: 6462134
#GC: 53.970000
Expects: FILE
Default: None
Input a genomic feature file to annotate the repeats in the genomic context. PERF accepts both GFF and GTF format genomic feature files. Each repeat is annotated w.r.t the closest gene and classified either as Genic, Exonic, Intronic and Intergenic according to the position of the repeat. Besides this, the repeat is also checked if it falls in the promoter region of the gene. Annotation adds 7 columns to the default perf output which already consist 8 columns.
S.No | Column | Description |
---|---|---|
9 | Gene name | Name of the closest gene |
10 | Gene Start | Start position of gene in the Chromosome |
11 | Gene Stop | End position of gene in the Chromosome |
12 | Strand | The strand orientation of the gene |
13 | Genomic annotation | Annotation of the repeat w.r.t to the gene. Possible annotations are {Genic, Exonic, Intronic, Intergenic} |
14 | Promoter annotation | If repeat falls in the promoter region of the closest gene. The default promoter region is 1Kb upstream and downstream of TSS. |
15 | Distance from TSS | Distance of the repeat from the TSS of the gene. |
Expects: STRING
Default: GFF
Option to specify the format of the input genomic feature file. Accepted inputs are GFF or GTF. More details about the GFF and GTF formats can be found here.
Expects: STRING
Default: gene
The attribute key used for the name of the gene in the GFF/GTF file. In the below example GFF file, we have the location of a gene and it's mRNA and exon locations. The last column of the file specifies attributes associated with each feature, like ID, Parent, gene etc. PERF uses on of the attribute to identify the gene and also it's exons. In th below example the key "gene" can be used to identify gene and the exons of the gene as they have the same gene name. Please check your GFF/GTF file for a robust attribute key which can identify all genes and their corresponding exons. We are actively working on better annotation where we can identify genes and their exons based on the ID and Parent.
# Sample GFF
NC_004354.4 RefSeq gene 124370 126714 . - . ID=gene1;Name=CG17636;gbkey=Gene;gene=CG17636;gene_biotype=protein_coding;gene_synonym=DmelCG17636,EG:23E12.1;
NC_004354.4 RefSeq mRNA 124370 126714 . - . ID=rna1;Parent=gene1;Name=NM_001103384.3;gbkey=mRNA;gene=CG17636;transcript_id=NM_001103384.3
NC_004354.4 RefSeq exon 126626 126714 . - . ID=id13;Parent=rna1;gbkey=mRNA;gene=CG17636;transcript_id=NM_001103384.3
NC_004354.4 RefSeq exon 125495 126259 . - . ID=id14;Parent=rna1;gbkey=mRNA;gene=CG17636;transcript_id=NM_001103384.3
Expects: INT
Default: 1000
Upstream distance(bp) from the TSS of the gene to be considered as promoter region. Default 1000.
Expects: INT
Default: 1000
Downstream distance(bp) from the TSS of the gene to be considered as promoter region. Default 1000.
Prints the version info of PERF.
The following examples assume that the file with input sequence in FASTA format is named my_seq.fa
.
# Find all monomer to hexamer repeats of >=12nt length
$ PERF -i my_seq.fa
# Specify output filename
$ PERF -i my_seq.fa -o PERF_output.tsv
# Specify fastq format
$ PERF -i my_seq.fastq --format fastq
# Find all monomer to hexamer repeats of >=12nt length and generate an HTML report
$ PERF -i my_seq.fa -a
# Specify output filename
$ PERF -i my_seq.fa -o PERF_out.tsv -a # HTML file is called PERF_out.html
# Find all monomer to hexamer repeats of >=12nt length and annotate them
$ PERF -i my_seq.fa -g my_seq.gff
# Specify feature file format and set downstream promoter region to 500bp
$ PERF -i my_seq.fa -g my_seq.gtf --anno-format gtf --down-promoter 500 # HTML file is called PERF_out.html
# Find all monomer to hexamer repeats of >=15nt length
$ PERF -i my_seq.fa -l 15
# Find SSRs with at least 6 repeating motifs (for all motif lengths)
$ PERF -i my_seq.fa -u 6
$ cat my_repeats.txt
AG AG 2 +
CT AG 2 -
GA AG 2 +
TC AG 2 -
# Find all AG repeats and generate an HTML report
$ PERF -i my_seq.fa -rep my_repeats.txt -a
# Ignore monomer and dimer repeats, and repeats with <4 repeating units
$ PERF -i my_seq.fa -m 3 -u 4
# Report only tetramer repeats of >=16nt length, and generate HTML report
$ PERF -i my_seq.fa -m 4 -M 4 -l 16 -a
In all the above examples, the output of PERF is saved to my_seq_perf.tsv
and the HTML report is saved to my_seq_perf.html
unless -o
is specified.
If you find PERF useful for your research, please cite it as follows:
PERF: an exhaustive algorithm for ultra-fast and efficient identification of microsatellites from large DNA sequences
Akshay Kumar Avvaru, Divya Tej Sowpati, Rakesh Kumar Mishra
Bioinformatics, , btx721
doi: 10.1093/bioinformatics/btx721
For queries or suggestions, please contact:
Divya Tej Sowpati - tej@ccmb.res.in
Akshay Kumar Avvaru - avvaru@ccmb.res.in