README.md
In seqMeta: Meta-Analysis of Region-Based Tests of Rare DNA Variants

seqMeta

Meta-Analysis of Region-Based Tests of Rare DNA Variants

Computes necessary information to meta analyze region-based tests for rare genetic variants (e.g. SKAT, T1) in individual studies, and performs meta analysis.

the latest released version from CRAN with

r install.packages("seqMeta")
the latest development version from github with

r if (packageVersion("devtools") < 1.6) { install.packages("devtools") } devtools::install_github("DavisBrian/seqMeta")

If you encounter a clear bug, please file a minimal reproducible example on github.

CRAN win-oldrel build error fixed.

Export c.seqMeta
Fixed version 1.4.2 of CompQuadForm changing the name of the result
Fixed canonical form CRAN url

Explicitly passed formula to check_format_skat. Previously relied on calling environment to define formula.
Return NA if standard error is 0 in burdenMeta and singlesnpMeta.
prepScores and prepScores2 scaled theta incorrectly when using a kinship matrix.
Minimum coxme version moved to 2.2-4.
prepCox function calculates the projection matrix incorrectly for collinear variants in the same gene.
Fixed impute_to_mean from producing an error in odd circumstances.
Added verbose functionality to prepScores2.
Added examples to prepScores2 documentation.
Explicitly import all functions in packages other than base to comply with new CRAN policy.

Migrated to git / github
- Bug Reports and Feature Requests should be submitted github.
Minimum R version moved to 3.1.0
Duplicated SNP in snpinfo gene no longer pulls from the genotype matrix twice.
Monomorphic snps with caf != 0 were handled incorrectly.
Binomial models when genotypes imputed outside of seqMeta did not match when models were imputed by seqMeta. Very slight differences in the covariance structure.
Replaced any(is.na(Z)) with anyNA(Z)
Range test now checks that genotypes are [0, 2].
SNPInfo in seqMetaExamples had incorrect type of snpNames and aggregateBy.
Automatically convert (with warning) aggregateBy and snpName columns to type character if they are not already.
Added new function prepScores2

prepScores2 is a drop in replacement for prepScores, prepScoresX and prepCox. The only difference is the family argument should be text. gaussian() becomes "gaussian", binomial() becomes "binomial" and "cox" is used for survival models. prepScores2 is much faster in cases where genotype imputation occurs.

enforces assumption that when a gene is being anlyzed the same snp can only be in that gene once (same snp can still be in multiple genes or in the same gene in SNPInfo)
Reorganized code
Moved data checks before imputation of genotypes
Moved other checks closer to where the data is calculated / used
Improved speed and reduced memory usage by
- Replacing genotype imputation apply code with a faster vectorized version.
- Replacing any(is.na(Z)) with anyNA(Z)
- Replacing range with min and max
- Replacing matrix operations of the form t(x)%*%y with crossprod(x, y)
- Removing duplicate calculations
- Removing orphaned code
- Removing second imputation of genotypes in the covariance calculation
- Converting genotype matrix of type data.frame to a matrix.
Added Error Messages for
- Missing weight for snp-gene pair
- Duplicate weights for snp-gene pair
- Non-numeric genotypes
- family != gaussian or binomial
Added Warning Messages for:
- Removing missing snps and genes from SNPInfo
- Removing duplicated snp-gene pairs