A python package for designing diagnostic CRISPR and other PCR-based diagnostic assays using whole genome data.

CRISPR diagnostic assays, such as SHERLOCK and DETECTR, require the design of a CRISPR guide RNA (crRNA) which contains a spacer sequence unique to the taxon of interest. Most assays also require the design of primers for nucleic acid pre-amplification, using techniques such as PCR, RPA, or LAMP. The increasing abundance of whole-genome sequence data can be leveraged to predict effective primers and diagnostic regions. This data typically takes the form of whole-genome sequences (FASTA files) or variant data from reads mapped to a reference (VCF files). The krisp package contains python functions and command line programs to infer diagnostic regions from either FASTA files or a VCF file plus a reference. The krisp_fasta command uses a k-mer based approach on FASTA input and the krisp_vcf uses a sliding-window-based analysis of variants in a VCF file. Both commands use optimized algorithms that take advantage of multiple CPUs and use minimal RAM.

krisp_fasta searches for diagnostic regions by converting input genome data into sorted kmer tables which are then intersected to find conserved and unique regions. krisp_fasta allows specification of the desired length of the diagnostic region as well as the lengths of the conserved regions both upstream and downstream of the diagnostic regions. This can be used to directly search for a diagnostic spacer sequence, e.g. a 28-nucleotide sequence which is conserved at every position except a single diagnostic position. Alternatively, one can search for conserved primer regions flanking a much larger diagnostic region, e.g. 60 nucleotide diagnostic regions flanked by 32 nucleotide conserved regions upstream and downstream. When conserved primer regions are included in the search, primers can be analyzed with Primer3 and results filtered by the presence of optimal primers.

krisp_vcf scans through a VCF file with the associated reference FASTA file to find clusters of one or more diagnostic variants. Data can be scanned to distinguish two or more groups of samples in a single execution of the command. A sliding-window analysis of the variants searches for locations where one or more variants unique and conserved to a group of samples are surrounded by conserved sequence or variants. Once such locations are identified, the conserved consensus sequence for the target group is inferred by applying the variants to the reference sequence. Primer3 can then be used to identify primers in the conserved flanking regions and the regions are sorted by the predicted quality of primers.

krisp requires python version 3.9 or greater and a UNIX-like operating system which provides the sort command line utility. Most operating systems have sort installed by default as it is required by the POSIX standard. To see if sort is already installed, run:

sort --help

If a help menu is displayed, then sort is installed. Various python packages are required as well, including pysam, Bio, primer3-py, and nltk. These should be installed automatically when krisp is installed.

krisp_fasta can be installed directly from the PyPI repository:

pip3 install krisp

Alternatively, krisp_fasta can be installed from source utilizing the Makefile in this repository. Simply run the following command in the directory with the downloaded source files:

make install

Similarly, krisp can be uninstalled by running:

make uninstall

or

pip3 uninstall krisp

The krisp package is written in python, and when installed, provides stand-alone command line utilities called krisp_fasta and krisp_vcf. All the functionality of krisp_fasta is provided through the command line and knowledge of python is not required. In all the examples below, krisp_fasta is executed through the command line.

In addition to krisp_fasta and krisp_vcf, installation also provides a second command line utility called kstream. kstream is a separate python program which extracts and parses kmers from an input FASTA file. krisp_fasta utilizes kstream internally. Users of krisp_fasta are not required to use kstream directly. See examples below on kstream usage.

All commands have help menus which can be activated by passing the -h or --help option to the program.

After the krisp python package has been installed, krisp_fasta should be available on the command line. The help menu for krisp_fasta can be accessed with the --help option on the command line:

krisp_fasta --help

In example 1, we will be directly searching for a spacer sequence for CAS13a LwaCas13a, which accepts a 28-nucleotide spacer sequence and has the highest accuracy when the diagnostic SNP is at the 3rd rightmost position. In this case, we are looking for a spacer which has the following form:

{25 conserved nts}{1 diagnostic nt}{2 conserved nts}

And we want the diagnostic nucleotide to be conserved in our "ingroup" and different in our "outgroup".

The general form of this command is as follows:

krisp_fasta {ingroup files} --outgroup {outgroup files} --conserved-left 25 --conserved-right 2 --diagnostic 1

If krisp_fasta was installed via the GitHub repository, then test files can be found in the "test_data/krisp_fasta" directory. Within the "test_data/krisp_fasta" directory, we can run:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1

The output you see should look something like:

left_seq,diag_seq,right_seq
CGACAAGATACTCTCGCAGCTTGGT,M,AG
TGACGCAGATCATCCCGCGCTTACT,K,A

This is in the CSV format, which can be easily read by spreadsheet programs and scripting languages like R for further processing. Instead of printing to standard output, the output can be saved to a file using the --out_csv option or using a bash redirect >:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1 --out_csv test_out.csv

or

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1 > test_out.csv

When verbose is enabled using the --verbose option, the files being used and progress updates are printed during runtime:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1 --cores 4 --verbose > test_out.csv

In this example, all output is written to the terminal with the verbose information preceding the alignments. Internally, the verbose data is written to stderr, whereas the alignments are written to stdout, so test_out.csv will not contain the progress messages.

A more human-readable alignment format can be output as well using the --out_align option:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1 --out_align test_align.txt

This produces a file with the following contents:

CGACAAGATACTCTCGCAGCTTGGTCAG : ingroup0
CGACAAGATACTCTCGCAGCTTGGTAAG : ingroup1
CGACAAGATACTCTCGCAGCTTGGTGAG : outgroup0;outgroup1;outgroup2
                        {#}

TGACGCAGATCATCCCGCGCTTACTGAC : ingroup0
TGACGCAGATCATCCCGCGCTTACTTAC : ingroup1
TGACGCAGATCATCCCGCGCTTACTCAC : outgroup0;outgroup1;outgroup2
                        {#}

We see that 2 potential spacer sequences were found. Each alignment is displayed as a reference sequence, followed by the alternative sequences with conserved nucleotides replaced with '.' if the --dot-alignment option is used:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 25 --conserved-right 2 --diagnostic 1 --dot-alignment --out_align test_align.txt

which produces:

CGACAAGATACTCTCGCAGCTTGGTCAG : ingroup0
.........................A.. : ingroup1
.........................G.. : outgroup0;outgroup1;outgroup2

TGACGCAGATCATCCCGCGCTTACTGAC : ingroup0
.........................T.. : ingroup1
.........................C.. : outgroup0;outgroup1;outgroup2

To the right of each sequence is the filename(s) in which the sequence was found, with multiple filenames implying that all of these files contain the same sequence. Note that it is possible to have multiple sequences listed for a single file, which may be common for repetitive regions or diploid (or higher) organisms.

krisp_fasta can search for conserved primer regions which span both the ingroup and outgroup and can use Primer3 to design primers. The general formula is the same as in example 1, except that our conserved and diagnostic region will be longer.

Let us assume that we are looking for an amplicon which is 100 nucleotides in total length, and contains at least 30 conserved nucleotides on each end where we can design primers. To find these regions, we can run the following krisp_fasta command:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved-left 30 --conserved-right 30 --diagnostic 40 --dot-alignment --primer3 --out_align test_align.txt

This command was designed to be deliberately similar to that in example 1. For convenience, krisp_fasta, when the conserved flanking regions are the same, we can simply specify --conserved 30. We can also specify the total amplicon length as opposed to calculating the diagnostic length: --amplicon 100. In which case, the above command is equivalent to:

krisp_fasta ingroup*.gz --outgroup outgroup*.gz --conserved 30 --amplicon 100 --dot-alignment --primer3 --out_align test_align.txt

Both commands should produce output like the following in "test_align.txt":

ACGCACAAGGACAAGTGCCACTAAACCAGCCAGCCCTGACGCAGATCATCCCGCGCTTACTGACCAAGCTGCGAGAGTATCTTGTCGATGGGAACGATAG : ingroup0
.............................................................T...................................... : ingroup1
.............................................................C...................................... : outgroup0;outgroup1;outgroup2
   └────────Forward─────────┘                                           └────────Reverse────────┘

Primer statistics:
 Direction  Penalty  Sequence                    Tm        Gc Percent  Self Any Th  Self End Th  Hairpin Th  Position Penalty  End Stability  Template Mispriming       Template Mispriming Th   
 Forward    7.74706  CACAAGGACAAGTGCCACTAAACCAG  64.24706  50.0        0.0          2.14676      0.0         0.0               4.0            -1.7976931348623157e 308  -1.7976931348623157e 308 
 Reverse    6.43757  TCGTTCCCATCGACAAGATACTCTC   61.93757  48.0        0.0          0.0          37.5163     0.0               3.2            -1.7976931348623157e 308  -1.7976931348623157e 308 

Pair statistics:
 Penalty   Compl Any Th  Compl End Th  Product Size  Product Tm  Product Tm Oligo Tm Diff  T Opt A   Template Mispriming      
 14.18463  0.0           0.0           94            84.32116    22.38359                  62.70608  -1.7976931348623157e 308 

In this example, krisp_fasta finds two regions which specify our input constraints, although only the first is shown in the output above. Note that these alignments are actually reverse complements. krisp_fasta assumes that the input genomes are double-stranded and creates the corresponding kmers for both strands, thus finding two solutions.

The addition of the --primer3 option causes Primer3 to be run for each potential region. When --primer3 is used, only regions for which it was possible to find primers are included in the output. Additionally, the Primer3 statistics are included in the alignment and CSV output.

Users are encouraged to use krisp_fasta in any way that they seem fit. In this example, we will be using krisp to find all 60-mers which are conserved across a group of input files. To do this, we will specify a conserved region of 30-nts in each direction and a diagnostic region of 0-nts.

krisp_fasta ingroup*.gz outgroup*.gz --conserved 30 --diagnostic 0                                        

The result of which is:

left_seq,diag_seq,right_seq
ACGCACAAGGACAAGTGCCACTAAACCAGC,,CAGCCCTGACGCAGATCATCCCGCGCTTAC
AGTAAGCGCGGGATGATCTGCGTCAGGGCT,,GGCTGGTTTAGTGGCACTTGTCCTTGTGCG
CGCACAAGGACAAGTGCCACTAAACCAGCC,,AGCCCTGACGCAGATCATCCCGCGCTTACT
GTAAGCGCGGGATGATCTGCGTCAGGGCTG,,GCTGGTTTAGTGGCACTTGTCCTTGTGCGT

In this case, there is actually no distinction between the ingroup and outgroup files as these are only required to differ in the diagnostic region which we specified to be of 0 length.

kstream is a flexible and powerful tool for extracting and parsing kmers from an input fasta file or input stream of sequences. Some available options are listed below, but see the help menu for a complete list.

--kmers allows the user to specify the length (or lengths) or kmers to extract.

--disallow allows the user to omit any kmers containing disallowed nucleotides. See --allow for its inverse.

--map-softmask will map all masked nucleotides (lowercase characters) back to uppercase characters in all kmers.

--omit-softmask will omit any kmer containing soft-masked nucleotides.

--complements will add the complementary kmers to the stream.

--canonicals will only print out the "canonical" kmers, e.g. those which come first alphabetically.

--sort will sort the resulting stream.

--parallel will utilize multiple cores to speed up extraction.

By default, kstream will parse a fasta file and print out every sequence if no kmer length is provided. For example:

kstream test_DNA.fasta

Will print every sequence entry in the input file and disregard all header names (lines starting with '>'). If an input file is not given, then kstream listens to the standard input pipe. The previous command can also be written as:

cat test_DNA.fasta | kstream

This example utilizes a few of the common command line arguments: For example, to extract all 6-mers which don't contain "N" or "n", in sorted order:

kstream test_DNA.fasta --kmers 6 --disallow "Nn" --sort

If we only wanted the "canonical" kmers, then we could add the --canonicals flag:

kstream test_DNA.fasta --kmers 6 --disallow "Nn" --sort --canonicals

or pass the previous stream into a new kstream instance:

kstream test_DNA.fasta --kmers 6 --disallow "Nn" --sort | kstream --canonicals

Both are equivalent but the former is likely to be faster when cpus are limiting.

After the krisp python package has been installed, krisp_vcf should be available on the command line. The help menu for krisp_vcf can be accessed with the --help option on the command line:

krisp_vcf --help

krisp_vcf requires VCF input (compressed or not) and the reference genome in FASTA format used to create it. It also requires a metadata CSV file with at least two columns: one containing sample IDs that match the column names in the VCF file and one with arbitrary values corresponding to which group each sample belongs to. The first few rows of the metadata file for the example data looks like this (extra spaces added to make it easier to read):

sample_id,group
9D1,NA1
BS2014-584,NA1
BS96,NA1
CC1008,EU1
CC1011,EU1
CC1033,EU1

Files used in the examples below can be found in the "test_data/krisp_vcf" of this repository. You can use them to test krisp_vcf using the example commands below. Here is a minimal command:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1

This will scan the variants in "variants.vcf" and look for diagnostic regions that distinguish samples labeled as "NA1", "NA2", or "EU1" from all other samples. The VCF file is provided with the option --vcf since it could also be piped from standard input. Which group each sample is assigned to is defined by the "metadata.csv" file. Since no output file was specified, CSV output is printed to the standard output and status updates are printed to standard error. When run on the command line, both the output (standard out) and the status updates (standard error) are printed to the screen, which is not ideal most of the time. We can save the CSV output to a file by redirecting the output using > or using the --out_csv option:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv

or

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 > test_out.csv

While the program is running it prints status updates like this:

Undiagnostic| Unconserved | No primers  | NA1         | NA2         | EU1         
20050       | 91          | 14          | 6           | 22          | 17          

These numbers are counts of variant clusters considered and why they were determined to not be diagnostic or which group they are diagnostic for.

The output of the program is saved in "test_out.csv", which has many columns:

• "group": The group the region is diagnostic for
• "chrom": The chromosome on the reference database
• "n_diag": The number of bases that are diagnostic in the region
• A group of columns that contain the reference positions of primers, the diagnostic regions, and nearby sequences
• A group of columns containing inferred consensus sequence for the target group for primers, the diagnostic regions, and nearby sequences
• A group of columns containing the output of Primer3

This format is ideal for post-processing with bioinformatic tools and programming languages like R.

A more human-readable output format can also be generated by supplying the --out_align option:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv --out_align test_align.txt

This prints each output as a plain text alignment with the Primer3 statistics:

## Phyram_PR-102_s0001:179343-179465 is diagnostic for EU1

Reference : ctgggatactactaccatttgaccaattcaaggccatgtatgcctacgaccgctactagcgttggaaagttg  T  cgcttgacttgtgccacagcaaatg  T  tgacatttgga  G  aattggtctccg
      EU1 : ........................................................................<C13>.........................<C12>...........<A12>............
      NA1 : ........................................................................  T6 .........................  T6 ...........  G6 ............
      NA2 : ........................................................................ T10 ......................... T10 ...........  G8 ............
   oligos : ctgggatactactaccatttgaccaattc                  gaccgctactagcgttggaaagttg  C  cg                         C  tgacatttgga  A  aattggtctccg
            └─────── Left primer ───────┘                  └─────────── crRNA ────────────┘                       └───────── Right primer ────────┘

Primer statistics:
 Direction  Penalty  Sequence                       Tm        Gc Percent  Self Any Th  Self End Th  Hairpin Th  Position Penalty  End Stability  Template Mispriming       Template Mispriming Th   
 Forward    2.04051  ctgggatactactaccatttgaccaattc  61.54051  41.37931    0.0          0.0          31.61709    0.0               2.17           -1.7976931348623157e 308  -1.7976931348623157e 308 
 Reverse    5.34751  cggagaccaattTtccaaatgtcaG      60.84751  44.0        0.0          0.0          0.0         0.0               3.51           -1.7976931348623157e 308  -1.7976931348623157e 308 

Pair statistics:
 Penalty  Compl Any Th  Compl End Th  Product Size  Product Tm  Product Tm Oligo Tm Diff  T Opt A   Template Mispriming      
 7.38802  5.70965       4.50833       123           83.24808    22.40056                  61.62791  -1.7976931348623157e 308 

This alignment output contains the reference sequence and the inferred consensus sequence for each group being distinguished. The alignment uses the following syntax:

• Dots (.): positions in the group consensus sequence that match the reference
• Lowercase letters: Sequence derived from the reference FASTA rather than variants from the VCF
• Uppercase letters in reference/primers: variants from the VCF, potentially conserved in all groups
• Series of uppercase letters with numbers (e.g. T6): A variant with the number of samples supporting that variant. For example, T6AG2 would mean 6 samples had a T and 2 samples had a AG insertion at this position.
• Series of uppercase letters with numbers flanked by angle brackets (e.g. <C12>): Variants diagnostic for the group they are printed in and the number of samples that have this variant.

Most of the commonly used settings of Primer3 can be set on the command line:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv --out_align test_align.txt --amp_size 50 100 --gc_clamp 3

Note how the addition of --amp_size 50 100 --gc_clamp 3 changes the size of amplicons returned and the kind of primers selected:

## Phyram_PR-102_s0001:182872-182966 is diagnostic for NA2

Reference : caatatacctcctcaaacatcacaccctcAacttcatttattgtcggCccaacgcatgG  T  CcgcttcaatccgcatactaagtatttcCaaccg A  
      EU1 : ........................................................... T14 ..................................G14 
      NA1 : ...........................................................  T6 ..................................<A6>
      NA2 : ...........................................................<G12>..................................G12 
   oligos : caatatacctcctcaaacatcacaccc       catttattgtcggcccaacgcatgg  G  cc        ccgcatactaagtatttccaaccg g  
            └────── Left primer ──────┘       └─────────── crRNA ────────────┘        └────── Right primer ──────┘

Other Primer3 settings that can be changed include primer length, annealing temperature, GC content, maximum temperature of secondary structures (e.g. primer dimers). See the help menu with krisp_vcf --help for more Primer3 options.

krisp_vcf can be made to run much faster by subsetting the input to specific chromosomes or positions of interest. To subset to a specific reference position (for all chromosomes), use the --pos option:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv --pos 100000 200000

To subset to a specific chromosome (a single sequence in the reference FASTA file, identified by its header), use the --chroms option:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv --chroms Phyram_PR-102_s0001

The --pos and --chroms options can be used together to subset to a specific position on a specific chromosome.

If multiple processors are available on the computer running krisp_vcf, parallel processing can be enabled by specifying the number of processors to use using the --cores option:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 NA2 EU1 --out_csv test_out.csv --cores 4

VCF files can contain erroneous variants depending on how they have been quality filtered. krisp_vcf can do its own quality filtering of variants, allowing unfiltered variants to be used as input. By default, variants must occur in 5 samples per group, be represented by at least 10 reads in each of those samples, and have a genotype quality score of 40. These defaults can be seen in the help menu displayed by krisp_vcf --help. Custom values can be set using the --min_samples, --min_reads, and --min_geno_qual options:

krisp_vcf metadata.csv reference.fasta --vcf variants.vcf --groups NA1 EU1 --out_csv test_out.csv --min_samples 3 --min_reads 30

This work is subject to the MIT License.

We would like to hear about users’ thoughts on the package and any errors they run into. Please report errors, questions or suggestions on the issues tab in this github repository. We also welcome contributions via a Github pull request.