Skip to content
This repository has been archived by the owner on Mar 17, 2023. It is now read-only.

Latest commit

 

History

History
299 lines (247 loc) · 13.2 KB

dashit.org

File metadata and controls

299 lines (247 loc) · 13.2 KB
Raw

DASHit: automated guide design

DASHit are programs for automated design of CRISPR guides. There are two main workflows,

  1. DASHit-reads: guide design from reads (e.g., from a FASTQ file), and,
  2. DASHit-seq: Guide design by specifying a target sequence.

Note: DASHit is installed and ready to use on a compute server. Please read DASHit server to learn how to connect. If necessary, we can also install DASHit locally on your laptops.

DASHit-reads: Guide design from reads (e.g., from a FASTQ file)

<<DASHit-reads>> DASHit-reads designs CRISPR guides by analyzing the output of an NGS run. Guides are chosen to hit the largest possible number of reads in the input.

The following runs through automated guide design on NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fastq, assuming that this file has been placed in /scratch/example/ on the DASHit server.

  1. Connect to dashit.czbiohub.org and enter the examples directory 
    cd /scratch/example/DASHit-reads/
  2. Convert input FASTQ -> FASTA 
    seqtk seq -A NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fastq > NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta
  3. Run crispr_sites to build a map of crispr sites->reads in which they appear 
    cat NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta | crispr_sites -r > NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_sites_to_reads.txt
    
    *Note*: The new option

    The format of NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_sites_to_reads.txt is

    AAAAAAAAAAAAAAAAAGAC    13109137 4177172 6275108 12688285 ...

    The first column is a crispr site that was found, the remaining numbers on the line are the indices of the reads in which that crispr site appears. The read index is just the order of the read in the input FASTA file.

  4. Run optimize_guides to find the list of guides that best covers the input 
    optimize_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_sites_to_reads.txt 100 1 > NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv

    The integer parameter 100 specifies how many guides to create, and the integer parameter 1 specifies how many times to attempt to cover each read with CRISPR sites.

    You can also optionally tell optimize_guides about the original input FASTA file in which case it will also output a random read hit by each site. This is useful to, for example, BLAST the reads that are being hit and sanity check that the guides are hitting what you think they’re hitting.

    optimize_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_sites_to_reads.txt 100 1 NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta > NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv

Scoring automated guide design

score_guides compares a list of guides against a FASTQ file and computes how many reads are hit by the guides. Continuing the example from above, we run

score_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta
Scoring guides against NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta hit 5239845/16617695 = 31.53%
Parsing FASTA took 11m8.358304821s

Note: Guide scoring currently takes a long time, e.g. in the above example scoring 100 guides against a 3.4 GB FASTA file took 11 minutes. If this becomes burdonsome, we can work on making it faster, either by adapting the crispr_sites code or using the _sites_to_reads.txt

Splitting a FASTA file

score_guides can split the input FASTA file, writing out the reads that were hit by guides and those that weren’t to two separate files. To do this, specify the -s option:

score_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta -s
Scoring guides against NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta hit 5239845/16617695 = 31.53%
Parsing FASTA took 11m14.83192505s
Wrote NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_dashed.fasta
Wrote NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_undashed.fasta

We can use score_guides to check that the DASHed and unDASHed FASTA files are correct:

score_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_dashed.fasta
Scoring guides against NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_dashed.fasta
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_dashed.fasta hit 5239845/5239845 = 100.00%
Parsing FASTA took 55.039976945s

score_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_undashed.fasta
Scoring guides against NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_undashed.fasta
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_undashed.fasta hit 0/11377850 = 0.00%
Parsing FASTA took 10m11.982396976s

Scoring many FASTA files at once

score_guides outputs some (hopefully) helpful progress and diagnostic output to stderr. Only a single summary line is output to stdout, though, so if you’re scoring many FASTA files, redirecting stdout will give you a nice output with one single line per FASTA, e.g.

for i in `ls *.fasta`; do score_guides NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv $i >> scoring; done
cat scoring
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_dashed.fasta hit 5239845/5239845 = 100.00%
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001.fasta hit 5239845/16617695 = 31.53%
100 guides in NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_guides.csv vs. NID0092_CSF_ATGTCAG-GGTAATC_L00C_R1_001_undashed.fasta hit 0/11377850 = 0.00%

DASHit-seq: Guide design by specifying a target sequence

<<DASHit-seq>> DASHit-seq designs CRISPR guides by trying to cover an input sequence with guides, subject to not spacing them too closely together or too far apart. Before being selected, the list of candidate guides is filtered down to remove guides for poor structural reasons and if they match a use specified list of off targets.

As an example, we re-run automated guide design on mouse RN45s, with off targets given by the mouse transcriptome (with RN45s removed).

  1. Connect to dashit.czbiohub.org and enter the examples directory 
    cd /scratch/example/DASHit-seq/
  2. Run DASHit-seq, specifying the input mouse RN45s sequence and the off target mouse transcriptome (with RN45s removed) 
    dashit-seq.py rn45s-long.fa --offtarget mouse_transcriptome_sans_rn45s > mouse_rn45s_guides.csv
  3. DASHit-seq will output the designed guides to the specified CSV file. The CSV file also contains some metadata and lists the guides that were filtered due to structural reasons 
    Designed CRISPR guides
TTGCTGCGGAGCATGTGGCT
CCCCAGTCAAACTCCCCACC
CTCCAACCGGCCGTCCCCGA
AACGAAACGAGACACGTGTG
TTCACCTTGGAGACCTGCTG
GGCAAGACAGTTACTGATAC
...
    
    CRISPR guides that were removed from consideration
CRISPR site, why it was excluded
AGAGAGGCGACGGAGGGGGG, homopolymer>5
GGTGGGTTCCCACGGGGCAC, hairpin:-----GTTCCC---GGGCAC
GACACTCGGGGGGCCGGCGG, gc_frequency; homopolymer>5
AAATGCACGCATCCCCCCCC, homopolymer>5; dinucleotide_repeats>3
TCCCCCCCCCAACCACCACA, homopolymer>5; dinucleotide_repeats>3
AAGACCCGAGCCCGGCGCGC, gc_frequency

The guide design from this example was previously done in a bespoke fashion by Josh. Josh’s guides are available in /scratch/example/DASHit-seq/RN45s-classic-guides.txt, so you can compare the guides DASHit-seq creates with the ones that Josh made.

BED file visualization

DASHit-seq will output a BED file indicating where the designed guides hit the input file. If you open the original input FASTA plus the BED file in a sequence viewer, for instance in IGV, you can visualize how the designed guides tile across the input sequence.

./bed.png

Controlling filtering

The parameters for offtarget filtering and poor structure filtering can be controlled from the command line. For detailed instructions on how to specify the filtering parameters, simply run

dashit-seq.py -h

You can control how precisely guides need to match off targets to be disqualified by the --offtarget_radius option:

--offtarget_radius OFFTARGET_RADIUS
		      Radius used for matching an off target. Specify this
		      as L_M_N which means filter a guide for hitting an off
		      target if L, M, N nucleotides in the first 5, 10 and
		      20 positions of the guide, respectively, match the off
		      target. e.g., 5_10_20 to require perfect matches;
		      5_9_18 to allow one mismatch in positions 6-10
		      positions and to allow 2 mismatches in the last 10
		      positions

To control how guides are disqualified for poor structure reasons, you can specify the following options

filtering options:
  these options control how guides are filtered for poor structure reasons

  --gc_freq_min GC_FREQ_MIN
			filter guide if # of Gs or Cs is strictly less than
			this number
  --gc_freq_max GC_FREQ_MAX
			filter guide if # of Gs or Cs is strictly greater than
			this number
  --homopolymer HOMOPOLYMER
			filter guide if strictly more than this number of a
			single consecutive nucleotide appears, e.g., AAAAA
  --dinucleotide_repeats DINUCLEOTIDE_REPEATS
			filter guide if strictly more than this number of a
			single dinucleotide repeats occur, e.g. ATATAT
  --hairpin_min_inner HAIRPIN_MIN_INNER
			filter guide if a hairpin occurs with >=this inner
			hairpin spacing, e.g., oooooIIIooooo, where the o are
			reverse complements and III is the inner hairpin
			spacing
  --hairpin_min_outer HAIRPIN_MIN_OUTER
			filter guide if a hairpin occurs with >=this outer
			hairpin width, e.g., oooooIIIooooo, where the o are
			reverse complements and ooooo is the outer hairpin

Generating an off target input file

The off target file specified for DASHit-seq is simply a list of CRISPR sites that occur in the off target sequence you want to avoid. This file can be prepared using crispr_sites in special_ops_crispr_tools.

  1. Convert the off target sequence into a FASTA file, say offtarget.fasta
    1. Run crispr_sites on the off target FASTA file 
      cat offtarget.fasta | crispr_sites > offtargets.txt

Note: Currently crispr_sites is also outputting additional information about each site which is used by DASHit-reads. You won’t need this information for use as offtarget, so you should open the offtargets.txt generated above and delete everything but the first column.

DASHit server

<<server>> David has setup dashit.czbiohub.org, an EC2 instance we can use for guide design that has all the DASHit software pre-installed. For now this makes things easier by not requiring you to install development packages on your local machines. If running DASHit remotely becomes too annoying, then talk to David about getting DASHit setup on your local machine.

Connecting

  1. Get an SSH key id_dashit from David or Emily, and place it in ~/.ssh/
  2. Set the permissions on the SSH key appropriately 
    chmod 600 ~/.ssh/id_dashit
  3. SSH in 
    ssh -i ~/.ssh/id_dashit dashit@dashit.czbiohub.org

Storage

The /home/dashit home directory is on a small 8GB partition used for the operating system. A 500 GB partition is mounted as /scratch, so use that for your FASTQ, etc files.

You can:

  1. Upload input files to /scratch directly, or,
  2. Run aws cp on the dashit server to copy directly from S3.
    • To do this you’ll need to setup your AWS credentials on dashit.czbiohub.org

DASHit source code

  • DASHit-seq is implemented as a literate programming file which you can view here here.
  • You can view the DASHit-reads programs, crispr_sites and optimize_guides, here and here.

Fresh install

  1. Install go 
    brew install go
    
    nix-env -i go