BYSTRO AI

Polygenic Risk Score Dashboard

Calculate ancestry-aware polygenic risk scores with automated variant QC, ancestry-matched allele frequency adjustment, LD-aware clumping, and continuous shrinkage. PRS analysis runs automatically after ancestry calculation.

Available Analyses

Built-in polygenic risk scores:

IBD — Inflammatory Bowel Disease Liu et al, 2015
Alzheimer's Disease — Late-onset Alzheimer's disease Jansen et al, 2019/Bellenguez et al, 2022

You can also add your own — see Custom Summary Statistics below.

What You'll See

Analysis Control Panel

One-click interface to start analysis. Select a disease, hit “Calculate PRS,” and watch the real-time progress indicator.

PRS Dashboard showing the Polygenic Risk Scores tab with Calculate PRS button and study selector

Risk Score Results

Ancestry-adjusted risk scores with percentiles calculated within your ancestry group. Odds ratios are converted to liability-scale betas and dosages are centered by ancestry-matched allele frequencies for accurate cross-population interpretation.

PRS Dashboard showing the Polygenic Risk Scores calculated for Alzheimer's Disease

Understanding Your Results

Important
PRS indicates genetic predisposition, not destiny. Lifestyle and environment matter significantly. Higher scores suggest increased genetic risk but do not guarantee disease development. Always interpret results with clinical context.

Reading the Table

PRS Value — 0 = average risk within your ancestry group. Positive = higher risk, negative = lower risk.
Odds Ratio (OR) — Ancestry-corrected risk vs. population average. OR = 1.0 is average, > 1.0 is increased, < 1.0 is decreased.
Percentile — Your position within your ancestry group — higher percentiles mean higher genetic predisposition.

Risk Levels

Based on ancestry-corrected odds ratios:

Protective / LowOR < 0.67
Below average predisposition, lower tail of distribution
AverageOR 0.67–1.49
Middle ~80% of the distribution
Moderate / ElevatedOR 1.5–2.9
Top decile, comparable to common modifiable risk factors
HighOR 3.0–4.9
Top 2–5%, clinically significant
Very HighOR ≥ 5.0
Top 0.5–1%, warrants clinical discussion
What to do with your results

Share with healthcare providers for personalized guidance.

Consider lifestyle modifications for higher-risk conditions.

Use to inform timing and frequency of medical screening.

Custom Summary Statistics

You can run PRS on virtually any trait by uploading GWAS summary statistics from the NHGRI-EBI GWAS Catalog. Thousands of studies provide harmonized files covering a wide range of diseases and traits. Download one and upload it to run through Bystro's full ancestry-aware PRS pipeline.

Finding Files in the GWAS Catalog

1

Search for your trait

Go to ebi.ac.uk/gwas and search for a disease or trait (e.g., “type 2 diabetes”, “breast cancer”).

2

Find a study with summary statistics

Not every study has downloads. Look for a summary statistics link on the study page. Larger, more recent studies are more likely to have one.

3

Download the harmonized file

The catalog provides both original and harmonized files. Download the harmonized version — it ends in h.tsv.gz.

4

Upload to Bystro

Upload the file as-is. No preprocessing needed.

Use harmonized files only
Bystro requires the harmonized version of summary statistics, not the original author-submitted files. Harmonized files have been standardized to a consistent format with genome coordinates mapped to a known reference. Non-harmonized files will cause errors or inaccurate results.

Upload Walkthrough

Click the upload icon on the PRS dashboard to go through the 3-step upload process:

1

Study Metadata

Label your study so you can identify it in your results.

Upload Custom Summary Statistics - Step 1: Study Metadata form

Study Name: e.g., “Jansen et al 2019” or “T2D Risk - Mahajan 2022”

Trait: Short code, e.g., “AD” or “T2D”

Human Readable Trait: Plain-language name shown in results, e.g., “Alzheimer's Disease”

2

Parameters

Analysis parameters for the PRS calculation. Defaults work well for most studies. Adjust only if you have specific requirements.

3

File Upload

Select ‘Choose hg19 file’ or ‘Choose hg38 file’ and hit submit. You need at least one build, but can upload both.

Upload Custom Summary Statistics - Step 3: File Upload
hg19 vs hg38?
Most recent studies provide harmonized files in hg38. Older studies may only have hg19. The GWAS Catalog download page indicates which builds are available — match the file to the right upload slot.

File Requirements

Format: TSV in GWAS Catalog Harmonized format (.tsv or .tsv.gz)

Max size: 5 GB per file

Methodology

Bystro's PRS uses algorithms developed by David Cutler at Emory University, with continuous shrinkage methods optimized for terabyte-scale datasets.

Liability-Scale Conversion

Converts odds ratios to frequency-independent liability-scale effects, producing accurate scores regardless of allele frequency differences between training and target populations.

Ancestry Integration

Incorporates ancestry inference results to select appropriate GWAS weights and population references for population-specific risk estimates.

Continuous Shrinkage & LD Clumping

Empirical-Bayes shrinkage with LD-aware variant clumping to reduce overfitting and intelligently weight variants across linkage disequilibrium blocks.