STAAR Matrix Dimension Error

[yl315504]

Hello,

We are interested in this tool, but got frustrated when running it.

Platform: Windows 11, R 4.2.0
STAAR Version: 0.9.8
Data: 1,225 samples, 561,187 variants (binary phenotype)

The Problem
When running STAAR analysis, I get this error:
Error: matrix multiplication: incompatible matrix dimensions: 1x1225 and 1000x859

My code:
library(STAAR)

Fit null model

null_model <- fit_null_glmmkin(
fixed = phenotype ~ Age + Menopause + HRT + PC1 + PC2 + PC3 + PC4 + PC5 + PC6 + PC7 + PC8 + PC9 + PC10,
data = analysis_data_complete,
id = "id",
kins = kins,
family = binomial(link="logit")
)

Run STAAR

result <- STAAR(
genotype = geno_filtered[1:1000, ],
obj_nullmodel = null_model,
annotation_phred = NULL,
rare_maf_cutoff = 0.01,
rv_num_cutoff = 2,
rv_num_cutoff_max = 1e+09
)

Diagnostics (All Pass)
✓ Genotype dimensions: 561,187 variants × 1,225 samples
✓ Null model samples: 1,225
✓ Sample IDs match between genotype and null model
✓ Missing values: <10%
✓ Null model converged: TRUE

Question
What does the "859" dimension represent in the error, and how can I resolve this?

Is this a known issue with STAAR on Windows/R 4.2.0?

[yuxinyuanqt]

Hi @yl315504 ,

Thanks for the detailed information.

The STAAR function expects genotype to be an n × p matrix (samples × variants). From your description:

561,187 variants × 1,225 samples

it appears that geno_filtered is likely p × n (variants × samples)?

In particular,

geno_filtered[1:1000, ]

subsets 1000 rows (variants), which suggests the matrix orientation may be reversed. This can lead to the reported dimension mismatch.

[yl315504]

It works now. Thank you so much!

I have a follow-up question. We have generated annotation results using FAVORannotator.
My question is that shall we use the generated GDS file from FAVOR as the input of STAAR annotation_phred?

Thanks,
Yan

[yuxinyuanqt]

Hi @yl315504 ,

Yes, once you have generated the annotated GDS (aGDS) file using FAVORannotator, you can use it as the source to extract both the genotype matrix for your mask of interest and the corresponding annotation_phred matrix for STAAR analysis.

In practice, the aGDS file stores both genotype information and functional annotations, and you would typically extract the variants belonging to the specific coding or noncoding mask you would like to analyze, together with their corresponding annotation scores.

You can refer to the implementation example here:
https://github.com/li-lab-genetics/STAARpipeline/blob/main/R/coding.R#L68-L181

This section shows how to extract the genotype matrix and annotation_phred from the aGDS file for downstream STAAR analysis.

You may also refer to our STAARpipeline R package and the STAARpipeline Tutorial for more details on the full workflow.

Hope this helps.

Best,
Yuxin