CRISPRme+ general indexes — wide-flank, PAM-less
One index that serves many PAMs. A general index stores the raw genomic
flank instead of a fixed PAM, so a single build answers searches for any 3′ PAM no
longer than that flank — NGG, NRG, NRR, NGA, NAG, NNGRRT and so on —
with the PAM applied at search time by --prune-pam (3′ PAMs only: see the
orientation limit below).
Pulled by crisprme.py download --what index --index-name <name>, which routes
to this repo automatically for any flank wider than 3.
Why this is a separate repo
These indexes are not backward compatible. A CRISPRme release that predates
--prune-pam reads exactly as many stored PAM bytes per leaf as the search PAM
is long. Handed a wider stored flank it reads the leaf stream at the wrong width,
desynchronises at the first leaf, and crashes — and because a crashed search used
to be reported as a clean empty result, that failure was silent.
CRISPRme+ 2.8.1 (CRISPRitz 2.9.1) refuses such a mismatch up front, naming
both widths, and resolves the width correctly under --prune-pam. Older clients
cannot, so these archives are kept out of
lucapinello/crisprme-data,
which older defaults point at. The split is derived from the flank width at
publish time rather than kept as a list of names, so it cannot be forgotten.
Minimum versions: CRISPRme+ 2.8.1 / CRISPRitz 2.9.1.
Layout
crisprme-data-general-index/
├── indexes/
│ ├── NNNNNNNN_3_hg38.tar.gz # reference-only, 8-nt flank
│ ├── NNNNNNNN_3_hg38+hg38_HGDP1kGP.tar.gz # HGDP + 1000 Genomes (4,091 samples, phased)
│ ├── genotypes_hg38_HGDP1kGP.tar.gz # per-sample genotype companion of the variant index
│ └── SHA256SUMS # checksum per archive
└── README.md
Indexes
| Index | Flank | Serves | Variants |
|---|---|---|---|
NNNNNNNN_3_hg38 |
8 nt | any 3′ PAM of length ≤ 8 | none (reference-only) |
NNNNNNNN_3_hg38+hg38_HGDP1kGP |
8 nt | any 3′ PAM of length ≤ 8 | HGDP + 1000 Genomes, genotyped-phased (SHAPEIT5 re-phased) |
The variant archive bundles its _INDELS companion. Its per-sample genotype
store, genotypes_hg38_HGDP1kGP.tar.gz, is published here as well, byte-identical
to the copy in the main repo: download fetches the genotype companion and the
reference index from the repo the index itself came from, so without a copy here
a per-sample search on this index would run with degraded Samples.
Orientation limit. These indexes serve a 3′ PAM only. They store the flank
on the 3′ side of the guide, so a 5′ PAM (Cas12a-style, such as TTTV) is not served
at any width: use a dedicated index, which crisprme.py build-index-only builds.
Usage
# the genome, annotations, PAMs and sample lists (plus the NRG_3_hg38 reference index)
crisprme.py download --what all --path .
# fetch the general index (routes here automatically on flank width)
crisprme.py download --what index --index-name NNNNNNNN_3_hg38+hg38_HGDP1kGP --path .
# then search with whatever PAM you need -- no rebuild
crisprme.py complete-search --genome Genomes/hg38 \
--vcf list_vcf.txt --samplesID list_samplesID.txt \
--guide guide.txt --pam PAMs/20bp-NRG-SpCas9.txt \
--mm 4 --bDNA 1 --bRNA 1 --prune-pam --index-path genome_library \
--annotation Annotations/dhs+encode_screenv4+gencode+cosmic.hg38.bed.gz \
--output my_search
Run the search from the folder you downloaded into: download writes
list_vcf.txt and list_samplesID.txt there, and --prune-pam needs
--index-path. guide.txt holds one spacer per line, padded with an N for each
PAM position (NNN for a 3-nt PAM).
--prune-pam is CLI-only: the web form does not pass the flag, so a general
index cannot be used from the web interface.
Equivalence claim
A PAM-less index of flank K searched with --prune-pam and a 3′ PAM of length
L ≤ K produces the same report as a dedicated index built for that PAM, but for
what a flank wider than the PAM cannot index at a contig's very end and start: a + strand
PAM with 1 to K − L nt after it to the end, and a − strand PAM in the first K − L nt,
are not indexed, and one exactly K − L nt from the start is reported on the + strand at
another position. On hg38 at K = 8 that is 145 + 101 NGG and 308 + 241 NRG sites, none
on a primary chromosome; CRISPRme+ says so when a search reads such a flank.
Measured genome-wide for guide ACTGAAATCTGTAAGCAGGC at NRG 4 mismatches + 1 DNA
bulge + 1 RNA bulge over HGDP + 1000 Genomes with per-sample resolution: the flank-8 arm, the flank-3 arm and the
dedicated NRG_3 arm give byte-identical integrated_results.tsv
files, 182,369 rows each (the dedicated arm's licence-gated COSMIC column, empty on its
fresh install, set aside); that guide has no site at a contig's end. The archives here
were built by CRISPRitz 2.9.1, which keeps a target whose flank holds an N: an index of
this flank built by an earlier release misses those sites, and CRISPRme+ 2.8.1 warns when
a search reads one.
Verifying a download
sha256sum -c --ignore-missing SHA256SUMS # from inside indexes/
License & citation
Released under the MGB Open Access License 1.0 (see license_link).
Commercial use requires a separate licence. Cite CRISPRme+ and the underlying
variant panels (HGDP, 1000 Genomes) as described in the
CRISPRme+ repository.
- Downloads last month
- 87