Outcome: this could not have been more successful
+ Unity Note: you cant make this work without people getting along Statisticians respected the domain Bench scientists respected statisticians Sequencing labs with respect for each other
+ We will become the resource for high quality study design Resource for reproducibility We are defining, tracking and managing interlab and intra- lab variability We will increase, ensure, promote reproducibility, accuracy and precision in micorbiome research....for both protocols and bioinformatics methods We are also demonstrating the efficacy of multi-institutional projects!
+ Outcomes from the MBQC-pilot We’ve now explored protocol space Labs chose very broadly from among nearly all possible protocol variables The HMP explored microbiome variation; we’ve outlined protocol variation All of that possible technical variation… …was no bigger than, and often smaller than, biological phenotypes And now we know exactly what its sources are: Extraction kit, amplification primer, sequencing platform, sequencing depth, sequence filtering, OTU clustering, OTU filtering Not any of the (many) other protocol differences!
+ Outcomes from the MBQC-pilot When a phenotype reproduces between labs, it’s particularly convincing Different sources of microbial samples (mocks) Different gut microbiome phenotypes (ICU) Within-lab variation was small (for QCed samples) Potentially large sources of variation weren’t Sample storage Many of the extraction options Sequencer chemistry Many of the bioinformatic software choices
+ Study design elements for MBQC-I from MBQC-pilot Systematically evaluate the same range of protocol variable Sample handling (overall fixed protocol) Extraction kits:MO-BIO, Qiagen, others? Homogenization:yes, no (fixed protocol) Barcoding strategy:single, dual (fixed protocols) Sequencer:MiSeq 250, 300, HiSeq, others? PhiX:0%, 1%, 5%, 10%, others? Bioinformatics Quality trimming:0, 5, 20 (fixed software, e.g. Trimmomatic) Quality filtering:0, 75% (fixed software) Stitching/assembly:6, 20 (fixed software, e.g. FLASH) OTU calling:Classification, clustering, mapping (varied software) Sample filtering:0, min. reads (fixed software) OTU filtering:0, singletons, 0.005%, 0.1% (fixed software) 7
+ Outcomes from MBQC-I Now we know what variables to be evaluated MBQC-I will provide explicit protocol recommendations of which values and choices are compatible We know what needs to be incorporated into negative/positive controls and standards MBQC-I will evaluate protocol results on these controls We know what applies to human gut stool samples MBQC-I will generalize these results to other human body sites and sample sources
+ Outcomes from the MBQC-pilot Bioinformatics Handling
+ Informatics Benchmarking Resources All raw reads/datasets Metadata Full documentation of informatics methods iPython scripts Teaching materials walk user through how to test for variability
+ Future - immediate Shakedown all participants for: Raw data Documentation for pipelines Other? Frame work for pipeline? 14
+ Study design elements for MBQC-I not included in the pilot Sample collection methods Modalities: kit, swab, etc. Environments: more than just stool Sequencing technologies NextSeq, HiSeq X, PacBio, others? Mock communities and standards Cell and DNA-based communities Representative of broader environmental range Kit controls (blanks) For multiple protocol stages 15
+ Next components to study Inclusion of additional body sites Inclusion of additional sampling methods Evaluation of mock communities and standardized samples Scope of study in samples, locations and participants Other 16