TOAST output
0
0
Entering edit mode
Sam ▴ 10
@sam-21502
Last seen 10 weeks ago
Jerusalem

In the TOAST package, the results include a beta term, beta_var,mu and effect_size.

  1. Does any of these terms have a biological interpretation?
  2. Is there a way to get the mean of the expression for the specific cell type (after the deconvolution) ?
library(TOAST)

data("RA_100samples")
Y_raw <- RA_100samples$Y_raw
Pheno <- RA_100samples$Pheno
Blood_ref <- RA_100samples$Blood_ref

outRF1 <- csDeconv(Y_raw, K=6, TotalIter = 1, bound_negative = TRUE) 

design <- Pheno[,"disease",drop=F]
design$disease <- factor(design$disease, levels=unique(design$disease))

props_vignette <- outRF1$estProp
colnames(props_vignette) <- colnames(Blood_ref)

Design_vignette <- makeDesign(design, props_vignette)

fitted_model_vignette <- fitModel(Design_vignette, as.matrix(Y_raw))





summary(res_table_vignette$Gran$effect_size)

head(res_table_vignette$Gran)

============Output===================

      Min.    1st Qu.     Median       Mean    3rd Qu.       Max. 
-28835.009     -0.566     -0.015     -9.812      0.499    184.234

                beta   beta_var          mu effect_size
cg16034991 1.1270472 0.08304893 -0.12320596    2.559623
cg13293535 0.8020964 0.04121726 -0.03557254    2.194664
cg01479768 1.3355404 0.11858170 -0.27028634    3.359986
cg15172529 0.8254852 0.05158534 -0.06037535    2.342684
cg11045746 1.1931990 0.11201747  0.01766526    1.942483
cg00414890 0.8940398 0.06705851 -0.01381030    2.063758
           f_statistics      p_value       fdr
cg16034991     15.29502 0.0001805514 0.2024207
cg13293535     15.14201 0.0001934513 0.2024207
cg01479768     15.04168 0.0002024207 0.2024207
cg15172529     13.20968 0.0004681206 0.3510904
cg11045746     12.70984 0.0005906279 0.3519388
cg00414890     11.91955 0.0008559312 0.3519388
TOAST • 889 views
ADD COMMENT
0
Entering edit mode

I have found out from CARseq article that

TOAST defines the effect size as β/(μ+β/2), where μ is base-line expression in one group, and β is the gene expression difference between two group.

ADD REPLY
0
Entering edit mode

So I understand that the mean across all conditions per cell-type is (μ+β/2). That is analogous for example to DESeq2's basemean. But

  • How come μ can be negative?
  • Is it recommended to take (μ+β)/μ as to get a feeling for what is the fold change after one got rid (somehow) of the negative values? I want filter for genes with a decent fold change.
ADD REPLY

Login before adding your answer.

Traffic: 991 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6