Genotype analysis
Reference-free imputation and PRS313 scoring
Results
Awaiting analysisYour analysis will appear here
Choose a genotype file and run the analysis to compare scores across five breast cancer phenotypes.
- Summary statistics for each phenotype
- Score distributions across simulation trials
- Trial-level data available as a CSV
Preparing analysis
Loading the model and reference files.
0 of 100 trials complete
Keep this tab open while the analysis runs.
| Phenotype | Mean | Median | SD | Min | Max |
|---|
Scroll horizontally for all statistics.
Score distribution
Variation across simulation trials
These are polygenic scores, not an absolute probability of cancer. Simulation spread reflects missing-input sampling; it is not a clinical confidence interval.
Methodology
What FastImpute calculates and how to read the output
Reference-free genotype imputation
FastImpute uses pretrained, chromosome-specific linear regression models to predict missing PRS313 variants from observed genotypes. Models were trained on 1000 Genomes data using input variants linked to the 23andMe v5 panel. Observed PRS313 dosages are retained when available.
Simulation trials
Missing input genotypes are sampled using the reference minor allele frequencies. Each trial repeats imputation and scoring. The table reports the mean, median, population standard deviation, minimum, and maximum across those trials. The distribution shows this sampling variation.
Five score outputs
The existing PRS313 scoring pipeline returns overall breast cancer, ER-positive, ER-negative, hybrid ER-positive, and hybrid ER-negative scores. ER refers to estrogen receptor status. Scores are weighted sums of allele dosages, not individual cancer probabilities or calibrated risk percentiles.
Data and privacy
The selected file is read in browser memory. Genetic data is not sent to an analysis server or saved by this page. Model weights, reference tables, and calculation libraries are downloaded when needed. Exporting results saves a CSV to your device.
Limitations
This is a research baseline for the 23andMe v5 panel. Performance can vary with ancestry, genotyping platform, missingness, and the training data. Simulation variability does not capture every source of model error. Results are not a diagnosis or clinical risk assessment.
Read the project methods and evaluation