For those of you eager to use T. thermophila SB210 MIC genome sequence
information, I offer this update and opportunity to provide feeback.
At the meeting in July, Ed summarized the state of the MIC genome project.
Our first batch of sequence was obtained as paired-end Sanger reads from an
8 kb insert MIC genome plasmid library, with the library and sequencing done
by the JGI through an application to the Community Sequencing Program. By
the time that JGI actually did the CSP project sequencing, they were phasing
out their Sanger sequencing - instead of the 3X genome coverage, the raw
reads represented less than would be necessary for a good assembly. To
supplement this sequence and also to complement the bacterial library
approach, Mary Couvillion in my lab made MAC and MIC genome libraries for
Solexa sequencing. We did 95bp reads of paired-end libraries (and shorter
reads of mate-pair libraries that are ultimately not being used). We did the
MAC as a control for the technology, which turns out to have been critical.
Since the July meeting in VT, three members of the bioinformatics core at UC
Davis have been devoting precious time to thoroughly analyze all the data
with rigorous quality control. This has been a highly recursive process,
because the quality of the filtered data impacts how to use it in mapping
and analysis, and then results from mapping impact how the data should be
filtered, and back and forth and back and forth. They are doing a stellar
job in return for relatively little compensation, as there isn't an official
source of funding for the project.
It turns out that much of the sequence data, both Sanger and Solexa, is poor
quality and/or uneven in distribution across the MAC genome. The high AT
content is problem, as recent publications from large genome centers attest
(Plasmodium has comparable AT and has been a challenge to sequence and
assemble). There are new Solexa methods that increase sequence quality and
help reduce the coverage bias, but for now we're concentrating on using the
data in hand.
The numbers for MIC genome size generated right before the meeting, without
the recursive process of filtering the data and evaluating the mapping
statistics, overestimated the amount of sequence that is likely to be
MIC-specific. It seems that the MIC has more MAC-destined DNA than
MIC-specific DNA. The strategy for candidate IES mapping has been evolving,
successfully. We are still deciding what to analyze about the candidate IES
sites as a large set - how much do we care about position relative to
predicted ORFs? We only nucleotide resolution for IES sites mapped from the
Sanger reads. The larger number of Solexa paired-end reads also defines IES
with some caveats, based on the false positive rate of IES detection in the
MAC library. There will be tracks in a genome browser that will let everyone
search for candidate sites based on increasing weight of evidence.
Before this gets too long, I need to come to the promise that I made to
several of you at the meeting, which was to make a sequence file of bona
fide MIC-specific DNA that one could use to design experiments. The Sanger
reads that have been filtered stringently for quality yet don't map to the
MAC are the best source of reliable MIC-specific sequence, but the volume of
sequence will be only a fraction of the total MIC-specific content (there is
much more sequence coverage in the Solexa reads, but we can't seem to filter
the junk out as reliably). As of yesterday, the plan is to generate one file
of Sanger reads with apparent MIC sequence content and another file with the
subset of apparent MIC sequences that can be LINKED to a region of the MAC.
This second file will also attempt to assemble IES scaffolds (this process
will be highly incomplete).
Here is the question, for anyone intending to use the MIC-specific sequence
file as soon as it is available: Is is more useful to have a few IES fully
assembled or to have a larger data set of thousands of partial IES linked at
some distance to a region of the MAC? Or is is most important to have the
IES junction sequences? Or is the MIC-specific sequence without linking
information adequate? I know all these things are interesting to analyze on
genome scale, but that's not within the scope of the current project. The
question is in the context of immediate use by the community for ongoing