trinculo1

Toolkit to carry out genetic association for multi-category phenotypes

Overview

Running a multinomial logistic analysis

Usage:

trinculo MODE: [--bfile GENOTYPES_PREFIX | --dosage GENOTYPES.DOSE ] [--out OUTPUT_PREFIX] --pheno PHENOTYPES.TXT --phenoname PHENOTYPE [--basepheno REF --missingpheno MISSING ] [--covar COVARIATES.TXT --normalize] [--condition SNPS.TXT] [--priors PRIORS.TXT | --defaultprior --priorsigma PRIORVAR --priorcor PRIORCOR --covarsigma PRIORVAR | --empiricalprior ] [ --select --phenops --errormat --fullmodel --waldstats ] [ --noassoc ] [--threads n] [ --precision DIGITS ] [ --fitIter NITER --fitAbsTol ABSTOL --fitRelTol RELTOL ] [--help]

Options

MODE - Either multinom or ordinal for multinomial or ordinal logistic regression

--bfile FILE_PREFIX - Genotype input in binary plink format (FILE_PREFIX.bed, FILE_PREFIX.bim and FILE_PREFIX.fam) --dosage FILE - Genotype input in dosage format --out PREFIX - The prefix used for output files [default is "trinculo"]

--pheno FILE - Phenotype input (same format as plink)--phenoname STRING -The phenotype in the phenotype file to analyse --basepheno STRING - The reference category to use for the multinomial analysis [defaults to first category encountered] --missingpheno STRING - The symbol to be use to indicate a missing phenotype in the phenotype file

--condition FILE - A file containing a list of SNPs (one per line) to conditon on (i.e. include as covariates)

--covar FILE - A file containing covariates to condition on (same format as plink) --normalize - Normalizes covariates to have a variance of 1 [off by default]

--priors FILE - A prior covariance matrix on effect sizes --defaultpriors - Construct a prior matrix based on default values --priorsigma VALUE - Set the prior variance on effect sizes within a disease[default is 0.04] --priorcor VALUE - Set the prior correlation on effect sizes across diseases [default is 0] --covarsigma VALUE - Prior on the covariate effect size [Default is 1] --empiricalprior - Calculate an empirical prior directly from the data --noassoc -do not do any association testing, just infer the prior

--select -Do Bayesian model selection (i.e. calculate marginal likelihoods for each sharing model) --phenops -Do a likelihood ratio test on each phenotype individually --waldstats -Calculate standard errors, Z scores and Wald p-values --errormat -Output the variance-covariance matrix for the log odds ratio --fullmodel -Output all parameter estimates and the full error matrix

--threads n Use up to n cores

--precision DIGITS Print output to DIGITS significant figures

--fitIter NITER Try up to NITER iterations when fitting models --fitAbsTol ABSTOL Terminate model fitting if results change by less than ABSTOL --fitRelTol RELTOL Terminate model fitting if results change by less than (RELTOL x value)

--help - Prints this message