lzamparo/crisprml

Library design for CRISPr screen

crisprml's Issues

Guide generation should be parallelized

Looks like this is a trivially parallel problem; each guide does not depend on the choices of any other guide, so I should be able to speed up the computation by data-parallelism across chromosomes

Assemble library candidates from results

We have 324561 exons split across 94838 transcripts.
We have 242027 unique exons to cover.
We have 1534276 guides that are specific enough for our purposes that cover those exons.

Now we have to choose a guide per exon to make up a feasibly sized library. There are far more guides than are required, and even then too many exons to cover even if we choose one guide / exon.

Will discuss with Turgut tomorrow, but for now I'll see how many exons we need to cover if we consider just the first two exons of each Tx.

Set up data on Lilac

• set up directory structure for tries, genomes, and output guides & figures.
• script to move genomes & tries from warm storage into compute storage
• recover mm10 coding exon region bed file [Done, below /warm/leslie/zamparol/crisprML]
• recover mm10 kmers file [Done, below /warm/leslie/zamparol/crisprML/kmers/mm10/Cas9]
• recover mm10, hg38 tries [Done, at /warm/leslie/pritykin/projects/CRISPR//_all/_all_trie.dat]

Need to recover the mismatch_score.pkl, pam_scores.pkl files

From the feature_generation_from_sequence.sh script, two of the inputs common to each job are

mismatch="/cbio/cllab/home/aperez/Scripts/Project_Scripts/Ventura_CRISPR/active/crispr-project/guidescan-crispr/guidescan/CFD_scoring/mismatch_score.pkl"
pam="/cbio/cllab/home/aperez/Scripts/Project_Scripts/Ventura_CRISPR/active/crispr-project/guidescan-crispr/guidescan/CFD_scoring/pam_scores.pkl"

Need to recover these files, either from Alex' storage on Hal, or Yuri's backup on Lilac

Scripts are hardcoded to work on TORQUE / MOAB cluster

Since I'll be running these on an LSF-based cluster, re-write scripts to conform to job submission standards

Run test job on one gene

• Get CDS for one mm10 gene
• Run scripts, see if sensible output produced

Screen guides to ensure they do not contain cut sites for restriction enzymes

These enzymes will be used for cloning either target into the contsruct for mouse genome wide screens, or the reporter:

BstX1
Blp1
BamH1