FlashFry

Note: If you used FlashFry before version 1.9, the command-line interface has changed slightly.

FlashFry is a fast and flexible command-line tool for characterizing large numbers of potential CRISPR target sequences. FlashFry can be used with any genome, and can run against non-traditional model organisms or transcriptomes. If you’re looking to characterize a smaller region or would like a nice web interface we recommend the GT-scan or crispor.org websites.

The easiest way to get started is to try the quick-start procedure and confirm that everything works on your system. The wiki contains more detailed tutorials for specific workflows. Thanks to @drivenbyentropy for the Java implementation of the ViennaRNA energy calculations.

Documentation

Quick start

First, make sure you are running Java 8 by typing java -version. From a Unix or macOS command line, download the latest FlashFry JAR:

wget https://github.com/mckennalab/FlashFry/releases/download/1.15/FlashFry-assembly-1.15.jar

Download and extract the sample data for human chromosome 22:

wget https://raw.githubusercontent.com/aaronmck/FlashFry/master/test_data/quickstart_data.tar.gz
tar xf quickstart_data.tar.gz

Build the database. This should take a few minutes (approximately 75 seconds on the original benchmark laptop):

mkdir tmp
java -Xmx4g -jar FlashFry-assembly-1.15.jar \
 index \
 --tmpLocation ./tmp \
 --database chr22_cas9ngg_database \
 --reference chr22.fa.gz \
 --enzyme spcas9ngg

Next, discover candidate targets and their potential off-targets. This example uses the EMX1 target with flanking sequence, which is required by on-target scoring models that evaluate sequence context:

java -Xmx4g -jar FlashFry-assembly-1.15.jar \
 discover \
 --database chr22_cas9ngg_database \
 --fasta EMX1_GAGTCCGAGCAGAAGAAGAAGGG.fasta \
 --output EMX1.output

Finally, score the discovered sites:

java -Xmx4g -jar FlashFry-assembly-1.15.jar \
 score \
 --input EMX1.output \
 --output EMX1.output.scored \
 --scoringMetrics doench2014ontarget,doench2016cfd,dangerous,hsu2013,minot \
 --database chr22_cas9ngg_database

The scored sites are written to EMX1.output.scored.

Nuclease profiles

Target geometry is defined by a nuclease profile: spacer length, PAM side, and one or more fixed-width IUPAC PAM patterns. FlashFry derives discovery regular expressions, spacer and PAM ranges, mismatch masks, and bin positions from that definition. The built-in profile IDs are:

cpf1-tttn-20, spcas9-ngg-nag-20, spcas9-ngg-20, spcas9-nag-20, spcas9-ngg-nag-19, and spcas9-ngg-19 (the original aliases and database IDs remain supported)
spg-ngn-20
sacas9-nngrrt-21 and sacas9-nngrrt-22
ascas12a-tttv-23 and lbcas12a-tttv-23
encas12a-tttn-23 and cas12a-2xnls-tttn-23

New databases use a version-2 header that stores the complete profile definition and target codec. Version-1 databases remain readable. Targets through 24 bases, including SpG, retain the original packed representation; longer targets, including the SaCas9 and 23-base Cas12a profiles, use the wide codec with an independent occurrence count. SpG is kept distinct from SpCas9, so SpCas9-calibrated scores are not applied automatically. The encas12a and cas12a2xnls scoring metrics use bundled 23-position CFD matrices and do not apply to AsCas12a or LbCas12a. See the nuclease profile developer guide for extension and compatibility details.

Custom fixed-width profiles

Custom fixed-width profiles can now be used for enzymes not covered by existing, pre-programmed sets (say, if you want a custom PAM with Cas9 or some other CRISPR variant that’s not yet supported). To use these you’ll need to make a profile for your enzyme; the details of which can be found in the Nuclease profiles document in the docs folder. Custom profiels can be checked with profile validate --profileFile PROFILE.properties and used to build a dedicated database with index --profileFile PROFILE.properties. The validated definition is embedded in the database header, so discovery and scoring do not require the profile file again. The requirements is that the PAM and spacer sequence combined (the matched pattern) is <= 32 bases long, as we use 64-bit words for fast lookup.

Cite

FlashFry is published in BMC Biology; if you find it useful please cite:

TY  - JOUR
AU  - McKenna, Aaron
AU  - Shendure, Jay
PY  - 2018
DA  - 2018/07/05
TI  - FlashFry: a fast and flexible tool for large-scale CRISPR target design
JO  - BMC Biology
SP  - 74
VL  - 16
IS  - 1
AB  - Genome-wide knockout studies, noncoding deletion scans, and other large-scale studies require a simple and lightweight framework that can quickly discover and score thousands of candidate CRISPR guides targeting an arbitrary DNA sequence. While several CRISPR web applications exist, there is a need for a high-throughput tool to rapidly discover and process hundreds of thousands of CRISPR targets.
SN  - 1741-7007
UR  - https://doi.org/10.1186/s12915-018-0545-0
DO  - 10.1186/s12915-018-0545-0
ID  - McKenna2018
ER  -