SPAmix and SPAmix+ methods

Main features

SPAmix and SPAmix+ are accurate and efficient approaches to associating complex traits (including but not limited to quantitative, binary, time-to-event, ordinal, and longitudinal traits) to single variant.

The two methods use restropective saddlepoint approximation (SPA) approaches and share the below features.

  • Model residuals (whose sum is zero) are needed as input. Users can select appropriate statistical models depending on the type of traits in analysis.
  • High computational efficiency to analyze large-scale biobank data with millions of individuals
  • High accuracy to analyze common, low-frequency, and rare variants, even if the phenotypic distribution (or residual distribution) is highly unbalanced (e.g. case-control imblance in case-control studies).

The two methods are different in terms of

  • SPAmix is designed to analyze an admixture population or multiple populations and can incorporate local ancestry information into analyses. The method is still only valid to analyze unrelated subjects.

  • SPAmix+ is a scalable, accurate, and universal analysis framework to control for population structure and family relatedness. The method is valid to analyze heterogeneous or admixed populations with falily relatedness.

Method Account for population structure Account for local ancestry Account for family relatedness Prospective/Retrospective
SPACox NO NO NO Prospective
SPAmix YES YES NO Retrospective
SPAGRM NO NO YES Retrospective
SPAmix+ YES YES YES Retrospective

Workflow of SPAmix

plot

The SPAmix paper and the SPAmix+ paper are two distinct papers

Please note that the SPAmix paper and the SPAmix+ paper are two distinct papers. Unfortunately, although the SPAmix algorithm was completed in 2021, its submission was delayed for some reasons. Consequently, the SPAmix+ algorithm will be presented in a separate paper rather than within the SPAmix paper.

  • SPAmix has been implemented in R package GRAB.

  • The package of SPAmix+ is in preparation.

Important notes about SPAmix in function GRAB.NullModel

  • If the left side of argument formula is model residual, please specify the argument traitType = "Residuals". Otherwise, the argument traitType can be specified based on the type of trait in analysis.

  • For method SPAmix, the top SNP-derived PC and related information is required in the arguments formula and control.

Quick Start-up Guide

The below gives examples to demonstrate the usage of SPAmix.

Step 1. Read in data and fit a null model 

library(GRAB)
PhenoFile = system.file("extdata", "simuPHENO.txt", package = "GRAB")
PhenoData = data.table::fread(PhenoFile, header = T)

SPAmix method also support both original trait or model residuals as input. For SPAmix, the confounding factors of SNP-derived PCs are required and should be specified in control.

library(GRAB)
PhenoFile = system.file("extdata", "simuPHENO.txt", package = "GRAB")
PhenoData = data.table::fread(PhenoFile, header = T)
N = nrow(PhenoData)
PhenoData = PhenoData %>% mutate(PC1 = rnorm(N), PC2 = rnorm(N))  # add two PCs
obj.SPAmix = GRAB.NullModel(Surv(SurvTime, SurvEvent)~AGE+GENDER+PC1+PC2, data = PhenoData, subjData = IID, method = "SPAmix", traitType = "time-to-event", control = list(PC_columns = "PC1,PC2"))

The same results can be obtained via using model residuals

obj.coxph = coxph(Surv(SurvTime, SurvEvent)~AGE+GENDER+PC1+PC2, data = PhenoData)
obj.SPAmix = GRAB.NullModel(obj.coxph$residuals~AGE+GENDER+PC1+PC2, data = PhenoData, subjData = IID, method = "SPAmix", traitType = "Residual", control = list(PC_columns = "PC1,PC2"))

Step 2. Conduct genome-wide association studies

For different types of traits and methods, the step 2 is the same as below.

GenoFile = system.file("extdata", "simuPLINK.bed", package = "GRAB")
OutputDir = system.file("results", package = "GRAB")
# step 2 for SPAmix method
OutputFile = paste0(OutputDir, "/Results_SPAmix.txt")
GRAB.Marker(obj.SPAmix, GenoFile = GenoFile, OutputFile = OutputFile)

Citation

  • SPAmix: A scalable, accurate, and universal analysis framework using individual-level allele frequency for large-scale genetic association studies in an admixed population (to be updated).

  • SPAmix+: to be submitted.