On the call –Sarah, Margie, Nick, Barbara, Alejandro, Paul, Jason, Derrek, Neda, Josh, Sudha, Lenore Launer … Arfan and Charlie could not be on call due to prior commitment but reviewed and sent around an analysis plan prior to call. Reinhold could not be on call. … Summary: Details of conversation are given further below: CHARGE and ENGIMA will each run GWAS of 3 phenotypes: 1) ICV 2) Total Brain Volume 3) HV Sample: Caucasian only at this phaseAll available in CHARGE and ENIGMA including additional samples since Nat Gen papers to be included in discovery. Use studies with HV and DNA alone for replication. Exclusions: ICV: None TBV: One analysis using same criteria as previously, another excluding stroke and dementia and ‘other neuro’- brain tumor, major head injury, craniectomy etc. Not excluding psychiatric illness such as depression, schiz, alcohol abuse. HV: One analysis excluding dementia (persons with MCI will not be excluded) and another including everyone. Platform: GWAS data imputed to latest release of 1000 G Phase 1 v3 , (March/April 2012)- ENIGMA will let us know their parameters as they complete QC, using only Caucasian panel- details below; timeline: October 12 Covariates: Age, Sex, study site and familial relationships + age2 and ageX sex interaction + 1st 4 principal components (done in ENIGMA) + in ENIGMA alone (dummy covariate for scanner). It was felt that in CHARGE this was taken care of by study site. Models to run: 1) ICV with above covariates 2) ICV with height as additional covariate(ENIGMA will check if all sites can do that, may be self-reported height recorded at time of MRI which is okay) 3) Total brain (does not matter if posterior fossa was included or not, use what study has) 4) Total brain adjusted for ICV as well 5) HV- mean of both sides (no adjustment for ICV) 6) HV –mean of both sides, adjusted for ICV Upload: Without filters, filter at meta-analysis. Perform meta-analysis at two sites to be able to cross-check and verify. Leads: Sarah and Alejandro, Arfan and Josh Authorship: alternate across two cohorts Timeline: next call in about 3 weeks, Jason/Derek will send Doodle, October for initial analysis Introduction Paul – Lets all look at Arfan’s protocol, are there any comments? Sudha – The idea here is to start our collaboration by doing a pooled Meta Analysis of ICV, Brain Volume and Hippocampal volume across both studies (CHARGE and ENIGMA) using 1000 genomes imputation. Paul – collaboration so far has been very successful, additionally Arfan and Charlie looked up more CHARGE hits in ENIGMA and that is promising http://enigma.loni.ucla.edu/enigma-vis/ Imputation Paul – Sarah where are ENIGMA sites with imputation to 1000G? o Sarah - People should all have it finished by the next week or 2 ENIGMA has been using the most recent version, while CHARGE has used the older version (Oct 2010) o Josh – CHS is close to completing imputation with latest version but will not have it next week, all core CHARGE sites have or are imputing. Time line to be determined for ASPS, 3C and TASCOG-likely ok by Oct. Sarah – changed a couple things in the imputation protocol o Removed everything that wasn’t a doubleton or higher o GIANT has done the same thing o 13-14 million usable SNV (autosomal) for Caucasians (out of ~42 million overall) o 23 million available online, 10 singletons in European template that were removed o The number depends on whether we are using the European or all template. o Also doing X Launer – what do we use for replication Paul – since the first ENIGMA, more sites have joined to do the second phase of analysis which includes all subcortical structures and we can add more sites than were initially in ENIGMA Sudha– ideally we can include everyone with 1000G imputation for the discovery sample o For replication we can look in samples with phenotype data and DNA but no GWAS or 1000G imputation Josh - don’t know what it will be like to combine 1000G data across releases but we did do that in the past- no final decision on this yet Sudha– smaller studies may be delayed on imputation, so about 2000 to 3000 samples o If we wait till end of Oct, everyone should have imputation, but not if we want to get started by September o Josh – do not see a real need to rush as anyone with the phenotype in the world is pretty much part of these 2 consortiums o Sarah – it is worth waiting till October Barbara – how many do we need for replication o Sudha– it will not be clear, ideally we want about the same sample size as the discovery but that will not be possible MAF Threshold Sudha– it may be sites upload everything and then we decide Diseases in populations Sudha– what do we do about people with dementia, MCI and psychiatric disorders? o Paul - Phenotypes measured differently – ICV ex Fairly routine step in ENIGMA was to have sites exclude people with disease o Sudha– excluded many people with dementia, included MCI, since not all studies had the same threshold for determining MCI o Would it be feasible to ask ENIGMA sites to remove AD Sarah yes, we can ask it should be ok o Jason - ENIGMA was done with only healthy and also with everyone Results were similar Pvals lower in larger sample Tests and covariates Sudha– some studies have a measure of ICV and some others don’t o CHARGE ran BV controlling for ICV, ENIGMA did not o Jason – HV was done controlling for ICV, Sudha– why was this done for HV and not TBV? CHARGE went back and forth decided not to for HV because it could introduce noise Jason – we did this a lot of ways, same effect size was given for each analyses o Sudha– need to decide on that between the 2 consortiums on ICV adjustment for HV and TBV Age may be an issue, with younger persons ICV adjsutment is less important perhaps We have a lot of older people so TBV/ICV was preferred as a measure of atrophy o Jason we are running with and without ICV Also age, sex, ageXsex, age^2, dummy variables for scanning differences, 4 MDS o CHARGE has site variable which takes care of scanning differences Derrek - ENIGMA2 doesn’t have BV o Sarah it wasn’t less interesting, just less powerful in ENIGMA1, as some sites did not have it. o We can try to get other sites to get it o Derrek we only have to ask new sites Jason – we are following up ENIGMA2 with additional tests including L/R Average, and separate M/F for analyses, would CHARGE be interested in this? o Sudha may be interested, will think about it Josh – what’s the logic? MF tests would be ½ the sample Sudha– there is some interest in laterality Pay penalties with all these multiple tests Paul - sex difference exploration was pushed by one site in ENIGMA Barbara – are we doing with or without controlling for ICV Sudha– We have done ICV controlling for height… o Jason ENIGMA does not have height o Height SNPS come up as top SNPS in ICV analyses o Height strongly correlated with ICV Margie- can we ask sites for height? Paul lets not commit to it at the moment but lets ask Authorship Paul – authorship, basically combine starred authors from CHARGE and ENIGMA with Sarah and Alejandro leading from ENIGMA Ethnicity Sudha– should we use Caucasians only? o Paul - there may be a political issue if we tell people we do not want their data o Sudha: African American cohorts do have phenotypes in the same way, can use for replication, exploration at lociConclusions Paul - Another call in a month or so before then encourage people for Oct finish date for imputations Josh if sites are quick with imputation we can use that to decide what thresholds to use