IHMC11 last session, Friday afternoon

Bin Hu (Beijing Genomics Institute): A new way to research into human gut microbiome.
Unfortunately he's so far only giving an overview of the general microbiome approach, and the history of less-efficient methods (cloning...).  HMP (6 body sites) and MetaHIT (Human Intestinal Tract) projects...

The BGI has 137 Illumina HiSeq sequencers and 27 SOLID 4 systems!  They have data for 425 people, investigating food, obesity, IBD, diabetes, HBV, other viruses.  But "We need more data."

(I had to move farther back to find a free place to plug my laptop in, and now I can't read the smallish print on his slides.)

Announces The International Earth Microbiome project!


Bruce Birren: "High-throughput 16S sequencing for human microgenomics"  But this talk seems to be about quality control - very interesting.  Generating the reference genome set for the HMP.

Clinical team established protocols to screen patients for 'normality' and for collecting biological material from them.  >17,000 primary specimens.  Converted these into DNA for the sequencing centers.

Consistency of results:  Controls: same samples to all 4 centers.  The center-to-center variation swamped the biological variation!  So needed to standardize the biological and informatics protocols.  The final protocols may not be the BEST, but they're the most reproducible.

Accuracy of results (this is 16S results from 454):  21 organisms, range of %G+C.  Yet another heat map.  Gave centers a mock-community.  Good correlation, but 5% of sequences weren't from the input organisms.  Problem turned out to be 454 chimeric reads.   Previously undetected because algorithms weren't sensitive enough. So developed a new tool: ChimeraSlayer.

Problem with classifying reads form different biological sites.  Not fixed yet but will be.  Problems with overestimating diversity:  13 input species gave >200 apparent species (OTUs)!

Data:  first 24 people:  Can we define a 'core' microbiome?  Plot percent of samples the organism is in, vs abundance of each organism.  The commonly-present organisms are not necessarily the common organisms.  How distinct are the microbiomes of the different body sites.  Many genera found in only one site (= specialist genera), but many present in all sites.  Specialist genera are not very abundant.

Overall, can't readily connect the 16S data to the biology.  Need not just more metasequencing but more sequences of reference genomes.

Karen Nelson (JCVI):  Tools for the trade.  Changed her talk to highlight gaps in info in this meeting.

Working groups and body-site specific working groups have been meeting every week for past 3 years, to generate list of organisms needed for scaffolding this kind of data, to serve as a community resource.

Original goal of 1000 reference genomes now 3000, also viruses and phage and protists.  Tree of 1400 species in the pipeline for genome sequencing.

NSF-funded study on role of diet and gut microbiota in non-human primates.

Claire Fraser-Liggett:  The role of gut microbiota in obesity in the Amish.  Metabolic syndrome an issue.  In general, the published papers haven't reached consistent conclusions.

Old-order Amish are a closed inbred community with well-documented ancestry (large families, 14 generations).  Diet quite consistent across families, not much snacking, not much prescription drug use.  GWAS data available for nearly 4000 people.

Lean, overweight, obese and metabolic-syndrome obese people, two samples > 3 months apart.  At level of bacterial family, can't tell the four groups apart.  Not at level of phylogenetic abundance.  Bacteroides:Firmicutes ratios differ but not significantly.  Very stable over time, maybe due to homogeneous Amish lifestyle.  No significant differences in diversity.  Household and heritability effects?  Not for household or siblings.

Core microbiome: maybe 13 genera., Bacteroides and Firmicutes???  A couple of species significant:  Ruminococcus, Faecalibacterium (marker of inflammation?)  11 genera linked to cholesterol and serum triglyceride levels and BMI.  See correlations between specific bacteria and the GWAS studies on the people carrying them.

So far they've only done 16S  RNA studies - showing who is there.  Next is to find out what they are doing.  Conclusion might be that gut microbiome is irrelevant to obesity (at least in OOA), but maybe the superficial similarity will turn out to be quite different when metagenome and transcriptome analysis is done.


George Weinstock:  What's next for the human microbiome (next meeting in Paris)?  He thinks the HMP has reached the stage where it's starting to generate testable hypotheses.  Momentum isn't flagging.

Compare to the human genome project:  He thinks there won't be a reference microbiome for each site.  Need to learn to distinguish 'healthy' microbiome variation and dysfunctional variation.  Need a multidimensional view (genes, transcripts, proteins, metabolites...)  Very valuable that we have close interactions between clinicians and researchers from the very beginning.

Coming hardware technology:  PacBio, 454 junior, Ion Torrent, MiSeq, SOLiD. 

PacBio: longer reads very valuable, run times shorter, needs less sample.  Example Enterococcus faecalis: ~50x coverage gives ~100% coverage.  Average read length 587 bases, 12% >1 kb.  Simply aligning the >2kb reads covers most of the genome.  Accuracy?  Peak 85% accuracy (= 15% errors) as advertised, verylow accuracy.  But errors are random, not systematic, so deeper sequencing corrects the consensus.  Use for 16S?  High error rate is a problem, but if can get whole 16S genes rather than little sub-sequences, could be very useful.

Computing technology:  continuing methods development.