gsMap Advanced Usage¶

Using Customized Latent Representations¶

# This could be any key in the obsm field of the AnnData object.
latent_customized = 'latent_customized'
gsmap run_latent_to_gene \
    --input_hdf5_path 'sample1.h5ad' \
    --workdir './workdir' \
    --sample_name 'sample1' \
    --annotation 'annotation' \
    --latent_representation $latent_customized

Conditional Analysis¶

Objective: Perform conditional analysis by adjusting for other functional annotations or cell-type-level annotations.

This step extends step 3: generate ldscore, by adding additional functional annotations to the baseline with the aim of conducting a conditional analysis. The directory of additional annotations can be specified using the parameter --additional_baseline_annotation. The other steps are same to the tutorials above.

Download the additional annotations:

wget https://yanglab.westlake.edu.cn/data/gsMap/gsMap_additional_annotation.tar.gz
tar -xvzf gsMap_additional_annotation.tar.gz

The format of the additional annotation files is such that each line represents a SNP, with columns indicating the annotation values for that SNP. These values can be either binary or continuous.

zless -S gsMap_additional_annotation/baseline.1.annot.gz

Execution: required memory: ~50G

for CHROM in {1..22}
do
    gsmap run_generate_ldscore \
        --workdir './example/Mouse_Embryo' \
        --sample_name 'E16.5_E1S1.MOSTA' \
        --chrom $CHROM \
        --bfile_root 'gsMap_resource/LD_Reference_Panel/1000G_EUR_Phase3_plink/1000G.EUR.QC' \
        --keep_snp_root 'gsMap_resource/LDSC_resource/hapmap3_snps/hm' \
        --gtf_annotation_file 'gsMap_resource/genome_annotation/gtf/gencode.v46lift37.basic.annotation.gtf' \
        --gene_window_size 50000 \
        --additional_baseline_annotation 'gsMap_additional_annotation'
done

gsMap on Biological Replicates¶

Objective: When multiple biological replicates are available, a uniform slice mean can be calculated for the gene ranks across the samples. This slice mean rank can then be used to compute the GSS. This approach ensures more consistent and comparable results across different samples.

Calculate the Slice Mean¶

To generate the slice mean, use the create_slice_mean command. This command outputs a parquet file containing the slice mean ranks for each gene. The --sample_name_list parameter specifies the names of the samples, and the --h5ad_list parameter provides the paths to the AnnData objects for each sample.

gsmap create_slice_mean \
    --sample_name_list 'E16.5_E1S1.MOSTA' 'E16.5_E2S11.MOSTA' \
    --h5ad_list 'gsMap_example_data/ST/E16.5_E1S1.MOSTA.h5ad' 'gsMap_example_data/ST/E16.5_E2S11.MOSTA.h5ad' \
    --slice_mean_output_file './workdir/sample_slice_mean.parquet' \
    --data_layer 'count' \
    --homolog_file 'gsMap_resource/homologs/mouse_human_homologs.txt'

Use the Slice Mean in Quick Mode¶

The quick_mode command allows you to run the entire pipeline in a single step using the slice mean. The --gM_slices parameter specifies the path to the slice mean file generated in the previous step.

# For the 'E16.5_E1S1.MOSTA' sample (the same applies to 'E16.5_E2S11.MOSTA')
gsmap quick_mode \
    --workdir './workdir' \
    --homolog_file 'gsMap_resource/homologs/mouse_human_homologs.txt' \
    --sample_name 'E16.5_E1S1.MOSTA' \
    --gsMap_resource_dir 'gsMap_resource' \
    --hdf5_path 'gsMap_example_data/ST/E16.5_E1S1.MOSTA.h5ad' \
    --annotation 'annotation' \
    --data_layer 'count' \
    --sumstats_file 'gsMap_example_data/GWAS/IQ_NG_2018.sumstats.gz' \
    --trait_name 'IQ' \
    --gM_slices './workdir/sample_slice_mean.parquet'

Use the Slice Mean in Step-by-Step Mode¶

To incorporate the slice mean into the step-by-step pipeline, provide the slice mean file using the --gM_slices parameter in the run_latent_to_gene command. This enables the computation of gene specificity scores based on the slice mean.

# For the 'E16.5_E1S1.MOSTA' sample (the same applies to 'E16.5_E2S11.MOSTA')
gsmap run_latent_to_gene \
    --workdir './workdir' \
    --sample_name 'E16.5_E1S1.MOSTA' \
    --annotation 'annotation' \
    --latent_representation 'latent_GVAE' \
    --num_neighbour 51 \
    --num_neighbour_spatial 201 \
    --homolog_file 'gsMap_resource/homologs/mouse_human_homologs.txt' \
    --gM_slices './workdir/sample_slice_mean.parquet'

Cauchy combination for multiple samples¶

Use the run_cauchy_combination command to aggregate the spot p-values for the same annotation across multiple samples.

gsmap run_cauchy_combination \
    --workdir './workdir' \
    --sample_name_list 'E16.5_E1S1.MOSTA' 'E16.5_E2S11.MOSTA' \
    --trait_name 'IQ' \
    --annotation 'annotation' \
    --output_file './workdir/combined_IQ_cauchy_combination.csv.gz'