Software suite developed for detection of chimeric or circular RNAs from LIGR-seq data (see citation below).
Sharma, E., Sterne-Weiler, T., O’Hanlon D., Blencowe, BJ. (2016) “Global Mapping of Human RNA-RNA Interactions.” Molecular Cell. 62(4):618-26
julia v0.4
perl v5 Packages
external software
ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/
ftp://ftp.ncbi.nlm.nih.gov/blast/db/
: nt, human_genomic, other_genomic optional
Install julia release here, which must be = v0.4 (v0.5 and v0.6 are not supported). If you are new to julia, or installing programs via command line, there is a helpful guide here. The packages can then be installed manually by opening the julia
REPL:
pkgs = ("ArgParse", "Match", "Distributions", "GZip")
map( Pkg.add, pkgs )
map( Pkg.test, pkgs )
The perl
packages can be installed using cpan -i Parallel::ForkManager
The current database is for hg19, you can download it here: hg19 transcriptome
It is also possible to use a different build or species, but automation of this process is still in development. In the meantime you should be able to reverse engineer the file formats in the hg19 version. Briefly fasta transcriptome headers need to be in the format of >ENST00001_ENSG00001_GENESYM_BIOTYPE
, for example >ENST00000432079.1_ENSG00000116747.8_TROVE2_protein-coding
. If you are having trouble feel free to open an issue or e-mail me.
Set up the default database like:
$ git clone https://github.com/timbitz/Aligater.git
$ cd Aligater
$ wget http://hgwdev.sdsc.edu/~timsw/GRCh37.v19.bt2.tar.gz
$ tar xzvf GRCh37.v19.bt2.tar.gz
The aligater executable is a wrapper for a set of tools that are meant to work as input/output for one another, often compatible with piping one into the next.
$ aligater -h
aligater [sub-command] [-h]
- align : align short-reads to transcriptome
- detect : detect chimeric reads by recursive chaining of transcriptome SAM blocks
- post : post-process LIG format files with BLAST or RACTIP
- reclass : create 2D k-mer db and reclassify chimeras using heirarchical type
- stats : compare crosslinked to mock-treated samples using multinomial statistics
- table : compile interaction results into tabular format
Using the default alignment parameters is recommended for LIGR-seq and the step can be run as simply as:
$ filename="somefile.fastq.gz"
$ nodir=`basename $filename`
$ prefastq=${nodir%%.*}
$ aligater align -x db $filename > sam/$prefastq.sam
or --bam
flag will pipe the output of aligater align into samtools view -bS
But feel free to alter other alignment parameters as you see fit for custom purposes:
$ aligater align -h
Example:
$ align_max=50
$ aligater align -p $cores -k $align_max --bam -x db/GRCh37.v19 $filename > bam/$prefastq.bam
The next step is detection, which can be piped from the first: aligater align | aligater detect > out.lig
$ detectparam='--gtf [annoFile.gtf(.gz)] --gfam [gene_fam.txt(.gz)] --rmsk [maskerFile.bed(.gz)]'
$ aligater detect $detectparam < sam/$prefastq.sam > lig/$prefastq.lig
for example with the default transcriptome you should have something like:
$ detectparam='--gtf db/GRCh37.v19.gtf.gz --gfam db/hg.gene_fam.txt.gz --rmsk db/GRCh37.repeatMasker.slim.bed.gz'
will print a lig formatted file to STDOUT.
or if bam output was specified from the first command.
$ samtools view -h bam/$prefastq.bam | aligater detect $detectparam > lig/$prefastq.lig
There are a few parts to the post processing step, a number of filtering flags, a --blast
(mandatory) step and a folding step using --ractip
(optional), each requiring the use of external dependencies (installed to your $path
), blastn
and ractip
, see requirements.
aligater post [filtering_flags] [--blast] [--ractip]
It is recommended that these commands be run separately, to effectively make use system resources.
For example, --ractip
takes several hours with substantial CPU on multiple cores (-p
), but doesn't require much RAM.
Meanwhile, --blast
is very CPU heavy and requires significant RAM as well!
The first command aligater post --blast
should be considered mandatory, it is not recommended to skip this step.
This step requires proper installation of blastn
and the blast databases to the environmental variable BLASTDB
.
$ export BLASTDB="/path/to/blast/databases"
$ blastn -version
blastn: 2.2.28+
Package: blast 2.2.28, build Mar 12 2013 16:52:31
$ aligater post [--tmp (def: /tmp/aligater)] [-p] --loose --blast < lig/$prefastq.lig > lig/$prefastq.blast.lig
Edit the tmp
directory to hit and number of threads -p
to use as necessary.
This is a two part step, you need to first create a junction library .jlz
from the
annotations of all samples and replicates that you plan to compare:
$ cat lig/*blast.filtered.lig | aligater reclass --uniq --save database.jlz
and then the actual reclassification step on each lig
file:
$ aligater reclass --load database.jlz -g -b -u < lig/$input.lig > lig/$input.reclass.lig
This step produces the final output of the aligater
package and takes a number of command line options that require that the input file names be properly formatted!
The input expects two foreground files, a crosslinking reagent treated (xlink) or mock-treated/control (unxlink) which are identical file names other than some specific identifier that you could specify as a wildcard. For example:
$ xlinkfile="sample_rep1-xlink.final.lig"
$ unxlinkfile="sample_rep1-unx.final.lig"
$ foregroundfile="sample_rep1-%.final.lig"
along with an --nd xlink,unx
flag which specifies the strings to insert into the %
wildcard to make the two input files.
Similarly, the two background files containing non-chimeric expression level .lig
files must be similarly formatted with a %
wildcard that will be interpolated with the same --nd
strings.
$ backgroundfile="sample_rep1-%.expression.lig"
$ nameParam="--fore $foregroundfile --back $backgroundfile --nd xlink,unx"
Additionally we have to supply the comma delimited normalization constants (total fastq input read numbers) --nc
for the foreground files:
$ totalAMTreads=73217814
$ totalMOCKreads=62648851
$ normParam="--nc $totalAMTreads,$totalMOCKreads"
This makes up the core arguments to aligater stats
:
$ aligater stats $nameParam $normParam > output.pvl
There are a number of other arguments which greatly expand the data compiled by aligater stats
such as the --vs
option which allows an arbitrary number of variable columns to summarize for each interaction 'col:type,col:type,etc..'. Each entry consists of a column number (1-based) followed by a :
and a data type character [pcfdn]
where each character stands for:
p
: paired column with colon delimited entries for example BIOTYPEA:BIOTYPEBs
: string containing column to includef
: floating point number within the scope of Float64d
: integer number within the scope of Int64n
: some member of the Number abstract type, could be complexFor default lig
format you can use: --vs 18:f,24:p,21:f,22:d,23:d
which should summarize most of the important columns.
NOTE: If you skipped the RactIP folding step in aligater post --ractip
then you will have a different number of columns than the default expects. To change command line options for aligater stats
to be compatible with your input .lig
files, use --gi 19
and --vs 18:p
.
Additionally there is a --filt
optional flag that can be used to specify a single column and regex pattern for filtering purposes. For example you may want to filter for only intermolecular interactions using --filt 1:I
.
The aligater detect
step outputs a basic .lig (tab delimited) format file, and the aligater post
outputs an extended version of the .lig fileformat. The standard format is:
Column | Example | Regex | Description |
---|---|---|---|
1 | I | [IPS] | Single letter code classification: I=Intermolecular, P=Putatively Paralogous, S=Intramolecular |
2 | A:B | [A-Z\:]+ | Chimeric Read Structure: A,B,C refer to molecules, : denotes a ligation. A:A = intramolecular ligation, A:B = intermolecular ligation, A:B:A = intermolecular ligation from A to B and then from B back to A |
3 | 1:3 | [\d\:]+ | Local alignments (from SAM) that have been chained to make up the chimera |
4 | RNU4-1:RNU6-1251P | \S+ | Gene symbols of transcripts aligned to chimeric read |
5 | ENSG00000200795.1:ENSG00000201372.1 | \S+ | Gene ids of chimeric segments |
6 | ENST00000363925.1:ENST00000364502.1 | \S+ | Transcript ids of chimeric segments |
7 | snRNA:snRNA | \S+ | Biotypes of transcripts in chimeric segments |
8 | U4_snRNA:U6_snRNA | \S+ or NA | If alignment overlaps RepeatMasker annotation, RepeatNames or NA are given in A:B order |
9 | snRNA:snRNA | \S+ or NA | If alignment overlaps RepeatMasker annotation, then RepeatClass or NA is given in A:B order |
10 | 68c3777d9e14bf33 | \S+ or [a-f1-9]+ | Readname or md5 hash of readname |
11 | ACTGGCAATTTAAAATTGGAA+ | [ATGCN]+ | Read Sequence, _ indicates ligation site |
12 | 124 | \d+ | Alignment Score |
13 | 1,30>30>17 | [\d\,>()]+ | Chimera uniqueness, deprecated |
14 | 74,26 | [\d,]+ | Alignment positions in transcripts |
15 | 74,27 | [\d,]+ or NA,NA | RepeatMask offset positions |
16 | 55,41 | [\d,]+ | Alignment lengths |
17 | chr12:120730966:-,chr20:42101679:+ | POS,POS where POS=\S+\:\d+\:[+-] | Genome start position of alignment segment |
The aligater stats
step outputs a .pvl
file consistent with Extended Data Table 1
of Sharma E, Sterne-Weiler T, et al. 2016:
Extended Data Table 1
Gene-ids : Comma delimited HUGO or Repeat family name in lexographical order
OE[+amt/-amt] : (+AMT/-AMT) / (Expected(+AMT)/Expected(-AMT))
AMT/Mock : (+AMT/-AMT)
Exp[AMT/Mock] : Expected(+AMT)/Expected(-AMT)
AMTReads : Number of reads in the +AMT sample + 1 pseudo count
OE-amt : Observed / Expected for +AMT sample (see Methods)
MockReads : Number of reads in the -AMT sample + 1 pseudo count
OE-mock : Observed / Expected for -AMT sample (see Methods)
AMT+ pval : Binomial p-value for significance in the +AMT sample
Mock pval : Binomial p-value for significance in the -AMT sample
AMT RPM : The transcript's expression in Reads per Million from -ligase sample
Mock RPM : The transcript's expression in Reads per Million from -ligase sample
Transcript Biotypes : Gene biotypes from GENCODE annotations
... any other summary line from --vs string