Hello Wei, The commands look right. To do some detective work, there are two things you could look in to, as a sanity check:
1) how much coverage are you getting between the genomes in the over.chain file? Many BLAT hits could be compressed into not-so-many chains. If you add up the coverage in the "to" database and "from" database, that might tell you something (if it is low for either one, then the low rate of liftOver would makes sense). 2) how long are the query regions in your "from" SB_Ste12_bind4.bed file? Have you tried to use shorter genomic regions representing a gene's footprint (BED6) or just the regions for exons from a particular transcript (BED12)? If you take those same regions - and run a BLAT against the "to" genome, are you able to capture hits that are not found with liftOver? Or do you find more that meet or exceed the original BLAT criteria? A few cases of each would be best. These kinds of tests can probably help you figure out where the problem is. I am also working on getting some feedback for the chain/net part of your process to see if there are some suggestions to offer. Thanks, Jennifer --------------------------------- Jennifer Jackson UCSC Genome Informatics Group http://genome.ucsc.edu/ On 5/3/10 11:52 AM, Wei Zheng wrote: > Hi, Jennifer, > > I get BLAT matches, but few liftOver matches (19 out of 256 regions). > Below are the code I used. > > cd /mnt/shared/GenomeLift/bayanus > #mkdir lift net psl chainRaw over bedOver > for i in {1..16} M; do faSplit -lift=lift/chr$i.lft size > ../sacCer1/chr$i.fa -oneFile 3000 /scratch/split/chr$i; done > for i in {1..16} M; do blat contig.fasta /scratch/split/chr$i.fa > raw/Sb_Sc_chr${i}.psl -tileSize=11 -minScore=100 -minIdentity=80 > -fastMap; done > cd raw; for i in {1..16} M; do liftUp -pslQ ../psl/chr$i.psl > ../lift/chr$i.lft warn Sb_Sc_chr${i}.psl; done > #faToTwoBit contig.fasta Sb_contig.2bit > cd ..; > for i in {1..16} M; do axtChain -linearGap=medium -psl psl/chr${i}.psl > Sb_contig.2bit ../sacCer1/chr${i}.2bit chainRaw/Sb2Sc_chr${i}.chain; done > chainMergeSort chainRaw/*.chain | chainSplit chain stdin > cd chain; for i in *.chain; do chainNet $i ../contig.sizes > ../../sacCer1/chrom.sizes ../net/${i}.net /dev/null; done > for i in *.chain; do netChainSubset ../net/$i.net <http://i.net> $i > ../over/$i; done > cat ../over/*.chain >../bedOver/over.chain > cd ..; > liftOver -minMatch=0.1 -multiple SB_Ste12_bind4.bed bedOver/over.chain > SB2SC_Ste12_bind.bed SB2SC_nomatch.txt > > > On Mon, May 3, 2010 at 2:31 PM, Jennifer Jackson <[email protected] > <mailto:[email protected]>> wrote: > > Hello Wei, > > > > To clarify, do you not have BLAT matches between the two species or > do you > > have the matches, but it is liftOver that not mapping data? If you don't > > know, try tweaking liftOver a bit first. > > > > liftOver parameters: > > - use BED as an input (if you were using positional format) > > - use -multiple, use -minMatch 0.1 > > - maybe add in a -minSizeQ 300 (or so) to keep short fragment out (that > > multiple will capture). Or leave out -minSizeQ, review, then add it > back in > > using a threshold you set based on what type of output you desire. > > > > Try the liftOver changes and let us know if this does not solve the > problem. > > If it doesn't, send back details for the processing you (your exact > > parameters based on the processes from the document) and we can provide > > feedback for loosening up the match criteria. > > > > Thanks, > > Jennifer > > > > --------------------------------- > > Jennifer Jackson > > UCSC Genome Informatics Group > > http://genome.ucsc.edu/ > > > > On 5/3/10 10:52 AM, Wei Zheng wrote: > >> > >> Hello, > >> > >> I was trying to generate over.chain from S. bayanus to S. cerevisiae > >> and perform liftOver to map TFBS of S.bayanus to S. cerevisiae > >> coordinates, using the instructions on > >> http://hgwdev.cse.ucsc.edu/~kent/src/unzipped/hg/doc/liftOver.txt. > >> However I can only get less than 10% of my TFBS lifted, the others are > >> always non-matched. > >> > >> I wonder whether you could point out some key parameters in blat, > >> axtChain, and liftOver steps for such closely related yeast species. > >> The assembly I used was sacCer1 (16 chromosomes downloaded from UCSC) > >> for S. cerevisiae and 1098 contigs (downloaded from SGD, reported in > >> Kellis 2003) for S. bayanus. > >> > >> Thank you very much! > >> > >> Wei > >> _______________________________________________ > >> Genome maillist - [email protected] > <mailto:[email protected]> > >> https://lists.soe.ucsc.edu/mailman/listinfo/genome > > > _______________________________________________ Genome maillist - [email protected] https://lists.soe.ucsc.edu/mailman/listinfo/genome
