PYRAMA: A Python tool for Robust Analysis and Meta-Analysis of Genome Wide Association Studies

---

Citation

If you use PYRAMA in your research, please consider citing:

Georgios A Manios, Sophia Nteli, Panagiota I Kontou, Pantelis G Bagos, PYRAMA: An open-source tool for advanced meta-analysis of genome wide association studies, Bioinformatics, 2026;, btag054, https://doi.org/10.1093/bioinformatics/btag054

Web Tool

The web tool available at: https://compgen.dib.uth.gr/PYRAMA/

📑 Table of Contents

• Features
• File inputs examples
  • 1. Discrete Phenotypes
  • 2. Continuous Phenotypes
  • 3. BETA/SE Format
• Requirements
• Installation
• Usage
  • Quality control
  • GWAS Meta-Analysis
    • Case: Direct BETA/SE (No Imputation)
    • Case: BETA/SE + Imputation
    • Case: Discrete Phenotypes (Standard, Fast, Bayesian)
    • Case: Continuous Phenotypes
• Plots
• Enrichment Analysis
• Use case scenario data
• Contributing
• License

Features

1. Quality Control
2. Robust analysis and meta-analysis of GWAS (For Discrete and Continuous phenotypes inputs)
3. Standard GWAS meta-analysis
4. Bayesian meta-analysis (For Discrete and Continuous phenotypes inputs)
5. Meta-analysis with summary statistics imputation (only for BETA/SE input)
6. Functional enrichment analysis

File inputs examples

1. Discrete Phenotypes

A tab-delimited file with the following columns:

SNP	CHR	BP	aa1	ab1	bb1	aa0	ab0	bb0
rs1	11	84095494	0	23	244	9	83	178
rs2	17	19683106	8	76	183	31	121	118
rs3	4	68129844	4	58	205	13	93	164
rs4	10	44481115	2	37	228	8	76	186
rs5	4	68116450	21	109	137	43	129	98

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located.
BP	Base-pair position of the SNP on the chromosome (genomic coordinate).
aa1	Count of individuals with the homozygous reference genotype in cases.
ab1	Count of individuals with the heterozygous genotype in cases.
bb1	Count of individuals with the homozygous alternate genotype in cases.
aa0	Count of individuals with the homozygous reference genotype in controls.
ab0	Count of individuals with the heterozygous genotype in controls.
bb0	Count of individuals with the homozygous alternate genotype in controls.

2. Continuous Phenotypes

A tab-delimited file with the following columns:

SNP	CHR	BP	xaa	sdaa	naa	xab	sdab	nab	xbb	sdbb	nbb
rs1000000	1	100	-0.02762	0.9914	283	-0.01343	0.9956	1866	0.0106	1.003	3102
rs1000000	1	100	-0.1341	1.053	45	-0.01535	0.9676	262	0.01574	1.013	514
rs1000000	1	100	-0.3615	1.053	10	0.06759	0.9186	93	-0.01571	1.040	170
rs1000000	1	100	-0.02187	0.9836	172	-0.002833	1.001	1320	0.00291	1.001	2578
rs10000010	1	200	-0.0109	1.001	1354	0.006966	0.9842	2628	-0.002799	1.032	126

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located.
BP	Base-pair position of the SNP on the chromosome.
xaa	Mean phenotype value for individuals with the homozygous reference genotype.
sdaa	Standard deviation of the phenotype in the homozygous reference group.
naa	Number of individuals with the homozygous reference genotype.
xab	Mean phenotype value for individuals with the heterozygous genotype.
sdab	Standard deviation of the phenotype in the heterozygous group.
nab	Number of individuals with the heterozygous genotype.
xbb	Mean phenotype value for individuals with the homozygous alternate genotype.
sdbb	Standard deviation of the phenotype in the homozygous alternate group.
nbb	Number of individuals with the homozygous alternate genotype.

3. BETA/SE Format

A tab-delimited file with these columns:

SNP	CHR	BP	A1	A2	BETA	SE
rs374450	1	12305285	A	G	-0.117157114	0.170367578
rs10489599	1	16585817	A	G	0.111312914	0.125714433
rs9725656	1	18460525	T	C	-0.038092709	0.123232082
rs1106839	1	19420391	A	G	0.004437813	0.124570835
rs2274001	1	19464772	T	C	0.012144373	0.124526074

Column	Description
SNP	SNP identifier (e.g., rsID)
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
A1	Effect allele (the allele associated with the reported effect size)
A2	Non-effect allele (the alternative allele)
BETA	Estimated effect size (OR/log(OR))
SE	Standard error of the estimated effect size

---

Requirements

• A Linux-based OS (e.g. Ubuntu 20.04 LTS)
• Python 3.10+
• Requirements specified in requirements.txt
• pyrama_beta_se_meta binary or script (for beta/SE meta-analysis), compile :

g++ -std=c++17 -O3 -pthread -o pyrama_beta_se_meta PYRAMA_beta_SE_meta.cpp

• 'ref' folder containing LD information of each reference panel in parquet file format (available for download here ). After downloading, extract to the working folder of PYRAMA. The Pheno Scanner LD reference panel is available upon request from the Pheno Scanner database (http://www.phenoscanner.medschl.cam.ac.uk)

---

Installation

1. Clone the repository:

   git clone https://github.com/pbagos/PYRAMA.git
   cd PYRAMA

2. (Optional) Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Ensure pred_ld.py is executable and on your PATH, and pyrama_beta_se_meta is compiled. Compile with the following command:

   g++ -std=c++17 -O3 -pthread -o pyrama_beta_se_meta PYRAMA_beta_SE_meta.cpp
   

5. If you wish to download the LD reference panels to conduct summary statistics imputation with PRED-LD, visit this link

---

Usage

Quality control

Before conducting an analysis or a meta-analysis, we recommend that users execute the quality control script that accompanies PYRAMA. This script performs essential preprocessing steps, including the removal of problematic rows from the input studies, verification of allele order (harmonization) and allele consistency across all datasets, and generation of a comprehensive quality control report. Additionally, it calculates the number of shared variants for each combination of studies, providing users with a clear overview of the variant over-lap among the imported studies.

python quality_control .py  --input_files study1.txt study2.txt [study3.txt ...]    --output final_merged_gwas.txt  [--skip_harm (optional)]

Flag	Description
--input_files	List of input GWAS files (tab-delimited). Each file should contain columns: SNP, CHR, BP, A1, A2, BETA, and SE. You must provide at least one input file.
--output	Output filename for the final merged and harmonized GWAS data. Will be written in tab-separated format.
--skip_harm	If set, skips the allele harmonization and filtering steps involving A1/A2. Use this when harmonization is not needed or when input lacks allele information.

GWAS Meta-Analysis

python pyrama.py  --i study1.txt study2.txt [study3.txt ...]  --o output_results.txt   [options]

Common options:

Flag	Description
--i, --input, --input_file	Paths to input data files (space-separated when multiple files are inserted)
--o, --output, --output_file	Path to output files
--inheritance_model	ADDITIVE, RECESSIVE, or DOMINANT
--effect_size_type	OR (odds ratio) or CATT (Cochran–Armitage trend test)
--robust_method	MIN, MAX, MERT, or FAST (MinP, Cauchy, MCM and CMC combination tests)
--type_of_effect	FIXED or RANDOM
--bayesian_meta	YES to run Bayesian meta-analysis (default NO). Available only for Discrete Phenotype Input
--imputation	Summary statistics imputation of missing SNPs (only for BETA/SE input)
--r2threshold	R² threshold for imputation
--population	Population code (EUR, AFR,EAS and SAS) for imputation
--maf	Minor allele frequency cutoff for imputation
--ref	Reference panel identifier (e.g. TOP_LD, Pheno_Scanner, Hap_Map, all_panels[default])
--missing_threshold	Fraction of studies that must include a SNP (default 0.5). When set to 0, it performs imputation in all given variants
-n, --nthreads	Number of parallel threads for file I/O (default: 1)
--het_est	Heterogeneity estimator: 'DL' (DerSimonian-Laird) [default], "'ANOVA' (Cochran-ANOVA), 'SJ' (Sidik-Jonkman)

Case: Direct BETA/SE (No Imputation)

python pyrama.py   --i study1.txt  study2.txt [study3.txt ...]   --o beta_se_meta.txt    

Case: BETA/SE + Imputation

python pyrama.py   --i study1.txt  study2.txt [study3.txt ...]   --o imputed_meta.txt    --imputation   --r2threshold 0.8   --population EUR   --maf 0.01   --ref all_panels   --missing_threshold 0.0    

Output columns

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
N	Number of valid studies for this variant
P	P-value of Random Effects meta-analysis
SE	Standard error
BETA	Random Effects estimate
I2	Heterogeneity index (I² statistic) in Random Effects meta-analysis
pQ	P-value for heterogeneity (often from Cochran's Q test)
BETA(FE)	Fixed Effect estimate
P(FE)	P-value for the Fixed Effect meta-analysis

Case: Discrete Phenotypes (Standard, Fast, Bayesian)

Flag	Description
--i, --input, --input_file	Paths to input data files (space-separated when multiple files are inserted)
--o, --output, --output_file	Path to output files
--inheritance_model	ADDITIVE, RECESSIVE, DOMINANT or ALL (to enable robust methods)
--effect_size_type	OR (Odds Ratio) or CATT (Cochran–Armitage trend test)
--robust_method	MIN (MIN2), MAX, MERT, or FAST (MinP, Cauchy, MCM and CMC combination tests)
--type_of_effect	FIXED or RANDOM
--bayesian_meta	YES to run Bayesian meta-analysis (default NO). Available for Discrete and Continuous Phenotypes Inputs
-n, --nthreads	Number of parallel threads for file I/O (default: 1)

• Standard

  python pyrama.py   --i study1.txt  study2.txt [study3.txt ...]    --o standard_meta.txt   --inheritance_model ADDITIVE   --effect_size_type OR   --type_of_effect FIXED  

• MIN2/MAX/MERT robust methods

  python pyrama.py   --i study1.txt  study2.txt [study3.txt ...]    --o standard_meta.txt   --inheritance_model ALL   --effect_size_type OR     --robust_method MIN --type_of_effect RANDOM  

Output columns

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
P	P-value of the analysis/meta-analysis
CI_lower	Lower confidence interval
CI_upper	Upper confidence interval
Weighted_Mean	Analysis estimate

• Fast Robust

  python pyrama.py     --i study1.txt  study2.txt [study3.txt ...]    --o fast_meta.txt     --inheritance_model ALL     --effect_size_type OR     --robust_method FAST   --het_est DL

Flag	Description
--i, --input, --input_file	Paths to input data files (space-separated when multiple files are inserted)
--o, --output, --output_file	Path to output files
--inheritance_model	ADDITIVE, RECESSIVE, or DOMINANT
--effect_size_type	OR (Odds Ratio) or CATT (Cochran–Armitage trend test)
--robust_method	MIN, MAX, MERT, or FAST (MinP, Cauchy, MCM and CMC combination tests)
--type_of_effect	FIXED or RANDOM
--bayesian_meta	YES to run Bayesian meta-analysis (default NO). Available for Discrete and Continuous Phenotypes Inputs
-n, --nthreads	Number of parallel threads for file I/O (default: 1)
--het_est	Heterogeneity estimator: 'DL' (DerSimonian-Laird) [default], "'ANOVA' (Cochran-ANOVA), 'SJ' (Sidik-Jonkman)

Output columns

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
Z_Dom	Z-score value of the dominant model
Z_Add	Z-score value of the additive (allelic) model
Z_Rec	Z-score value of the recessive model
P_Dom	p-value from the Dominant model
P_Add	p-value from the Additive (Allelic) model
P_Rec	p-value from the Recessive model
P_MinP	p-value from the MinP combination test
P_CCT	p-value from the Cauchy combination test (CCT)
P_CMC	p-value from the CMC combination test
P_MCM	p-value from the MCM combination test
P_Stouffer	p-value from the Stouffer's method for dependent tests combination test

• Bayesian

  python pyrama.py     --i study1.txt  study2.txt [study3.txt ...]     --o bayes_meta.txt     --bayesian_meta YES     --inheritance_model ADDITIVE     --effect_size_type OR     

Output columns

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
P	P-value of the analysis/meta-analysis
Z	Z-score value of the analysis/meta-analysis
CI_low	Lower confidence interval
CI_upp	Upper confidence interval
E_m	Posterior expectation for the population parameter μ
E_tau_square	Posterior expectation of the between study variability τ^2
V_mu	Variance of the posterior distribution for the population parameter μ
V_tau_square	Posterior variance of the between study variability τ^2

Case: Continuous Phenotypes

python pyrama.py   --i study1.txt  study2.txt [study3.txt ...]    --o continuous_meta.txt   --inheritance_model ADDITIVE   --robust_method MAX   --type_of_effect RANDOM

Flag	Description
--i, --input, --input_file	Paths to input data files (space-separated when multiple files are inserted)
--o, --output, --output_file	Path to output files
--inheritance_model	ADDITIVE, RECESSIVE, or DOMINANT
--effect_size_type	OR (Odds Ratio) or CATT (Cochran–Armitage trend test)
--robust_method	MIN, MAX, MERT, or FAST (MinP, Cauchy, MCM and CMC combination tests)
--type_of_effect	FIXED or RANDOM
--bayesian_meta	YES to run Bayesian meta-analysis (default NO). Available for Discrete and Continuous Phenotypes Inputs
-n, --nthreads	Number of parallel threads for file I/O (default: 1)
--het_est	Heterogeneity estimator: 'DL' (DerSimonian-Laird) [default], "'ANOVA' (Cochran-ANOVA), 'SJ' (Sidik-Jonkman)

Output columns

Column	Description
SNP	SNP identifier (e.g., rsID).
CHR	Chromosome number where the SNP is located
BP	Base-pair position of the SNP on the chromosome
P	P-value of the analysis/meta-analysis
Z	Z-score value of the analysis/meta-analysis
CI_lower	Lower confidence interval
CI_upper	Upper confidence interval
Weighted_Mean	Analysis estimate

Plots

You can generate Manhattan plots and QQ-plots from your GWAS analysis/meta-analysis results.

python plots.py   --i  results_file.txt

Arguments	Description
--i, --input	Path to input tab separated file with SNP,CHR,BP,P columns
--o, --output	Prefix for output files (default: gwas_plot)
--dpi	Resolution for figures in dpi (default: 600)

Enrichment Analysis

You can perform Enrichment Analysis with gProfiler for your GWAS analysis/meta-analysis results.

python enrichment_analysis.py -i input_snps.txt -o gprofiler_results.csv -p gprofiler_manhattan.png -s 0.01 --organism hsapiens --sources GO:BP KEGG REAC --no_iea

Arguments	Description
-i, --input	Input tab-delimited file with 'SNP' and 'P' columns
-o, --output	Output CSV file for enrichment results
-p, --plot	Output PNG file for Enrichment Analysis Manhattan plot
-s, --significance	Adjusted p-value significance threshold [default=0.05]
--organism	Organism name for g:Profiler (e.g., 'hsapiens')
--sources	Data sources to query (e.g., GO:BP GO:CC KEGG). If omitted, all available sources will be used.
--no_iea	Exclude electronic GO annotations (IEA)

Use case scenario data

The replication report is available at analysis_code.html file of this repository with all data and results.

Contributing

Contributions welcome! Please:

1. Fork this repo
2. Create a feature branch
3. Open a pull request with tests/examples

---

License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for details.