Dinh/Dinh 2012/NOTES/2012-2-18
Jump to navigation
Jump to search
Demethylation of neuronal LINE repeats and consequences on total RNA[edit]
- Download RNA-seq data from http://www.ncbi.nlm.nih.gov/projects/geo/query/acc.cgi?acc=GSE21161
Use FTP site: ftp://ftp-trace.ncbi.nlm.nih.gov/sra/sra-instant/reads/ByStudy/sra/SRP/SRP002/SRP002303 SRP002303 is RNA-seq SRP002304 is ChIP-seq
- Download and install SRA Toolkit on genemapster:
wget http://ftp-private.ncbi.nlm.nih.gov/sra/sdk/2.1.9/sra_sdk-2.1.9.tar.gz decompressed with tar -xzvf and compiled with 'make OUTDIR="/usr/local/sra/"' software files are in: /usr/local/sra/linux/rel/gcc/x86_64/bin
- abi-dump
./abi-dump ./abi-dump: error while loading shared libraries: libsradb.so.3: cannot open shared... ADDED LIBRARY TO PATH: export LD_LIBRARY_PATH=/usr/local/sra/linux/rel/gcc/x86_64/lib:$LD_LIBRARY_PATH Now abi-dump works!
- GO.wget.sh
for f in SRR039864 SRR039865 SRR039866 SRR039867 SRR039868 SRR039869 SRR039870 SRR039871 SRR039872 SRR039873 SRR039874 SRR039875 SRR039876 SRR039877 do # wget ftp://ftp-trace.ncbi.nlm.nih.gov/sra/sra-instant/reads/ByStudy/sra/SRP/SRP002/SRP002303/$f/$f.sra /usr/local/sra/linux/rel/gcc/x86_64/bin/abi-dump -A $f -M 30 -O Neurons_mouse_RNAseq $f.sra done;
Bowtie[edit]
- Download colorspace index: UCSC mm9 -
wget ftp://ftp.cbcb.umd.edu/pub/data/bowtie_indexes/mm9_c.ebwt.zip
- Decided to use Bowtie to map, report reads with multi-hits (-k 2), and then use the decoded sequences to BLAST to RepBase
--max option can't output decoded sequences.
- Reasons why you can't easily/confidently convert Colorspace to basespace: http://seqanswers.com/forums/showthread.php?t=16115
- In summary, if there is an error in the colorspace, it will lead to many more error in the conversion to basespace. It is better to perform the alignment, and each call leading to SNPs will not be trusted, and the mapping algorithm can better get the correct decoded sequence using dynamic programming.
- Bowtie command:
/bowtie -C -n 2 --best --strata -l 28 -k 2 -p 8 BowtieIndex/mm9_c -f SRR039864_F3.csfasta -Q SRR039864_F3_QV.qual KCl_hr0_b1_a.bowtie.map 2> KCl_hr0_b1_a.log
- When parsing the output -
Take as multi-hit reads which have two reported alignment. But multi-hit reads which disagreement between the decoded sequence will be tossed (low quality mapping and sequence). File:Bowtiemap2multihitfasta.txt
Repeat mapping with BLAST![edit]
- I have to rebuild the database because meangenemachine is not working.
- The problem I had previously was that >50% of multi-hit reads couldn't be map to the RepBase.
*Reads that map to pseudo-genes or genome duplications may be considered as multi-hit reads (does to uniqueness requirement) so this higher percentage of multi-hit reads not mapping to RepBase may be expected.
- New strategy
Download the latest RepBase update (2-03-2012) Merge all mouse relevant repeat files: less rodrep.ref rodsub.ref mousub.ref appendix/rodapp.ref | sed 's/\t/#/g' > mod.rodrep_all.ref NOTE: rodsub.ref, mousub.ref, apprendix/rodapp.ref are very small, so their addition should not make the database too large or let in multiplicity of repeat sequences.
- formatdb
formatdb -p F -i mod.rodrep_all.ref -l mod.rodrep_all.ref.formatdb.log -V -t RodentRepbase
- makembindex
makembindex -input mod.rodrep_all.ref -nmer 8 -logfile mod.rodrep_all.ref_makembindex.log -output mod.rodrep_all.mbindex
- blastn 2.25+ options to consider:
-query <File_In> -task 'megablast' (default, faster than blastn, uses a seed index table) -db <db name> -out <File_Out> Fine-tuning: -evalue 0.0001 -use_index 'true' (for MegaBlast use) -index_name <string> (for MegaBlast) -outfmt 6 (tabular) -perc_identidy 0.9 -culling_limit 1 (should report only the best hits) -ungapped -num_thread 8
- Finally I decided to use -task blastn since our database is not very big (Megablast wasn't helpful.)
blastn -task blastn -query $fafile -out mouseNeurons_$f.hits -db $dbfile -evalue 0.0001 -word_size 11 -outfmt 6 -perc_identity 0.9 -culling_limit 1 -ungapped -num_threads 8
- When I go to top, it doesn't show that 800% CPU is being used, instead only 60%. However, the output files are being generated quite fast.
- I am performing this BLAST search for ALL of our previously analyzed dataset, using 1 million multi-hit reads for each.
24 mouse neurons dIP+ChIP set 4 mouse ESCs dIP set 3 mouse neurons total RNA-seq (0hr, 1hr, 6hr KCl treatment).