Related Concept Videos
You might also read
Related Articles
Articles linked to this work by shared authors, journal, and citation graph.
Sort by
Same author
Genome-wide association and Mendelian randomization analyses link <i>Helicobacter pylori</i> infection to Human Leukocyte Antigen polymorphisms and autoimmune diseases.
medRxiv : the preprint server for health sciences·2026
Same author
JG2: an updated version of the Japanese population-specific reference genome.
Human genome variation·2025
Related Experiment Video
Updated: Jun 23, 2025

09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.0K
Two-stage strategy using denoising autoencoders for robust reference-free genotype imputation with missing input
Kaname Kojima1, Shu Tadaka2, Yasunobu Okamura2,3
1Tohoku Medical Megabank Organization, Tohoku University, 2-1 Seiryo-machi, Aoba-ku, Sendai, Miyagi, 980-8573, Japan. kojima@megabank.tohoku.ac.jp.
Journal of Human Genetics
|June 25, 2024
Summary
Deep learning offers reference-free genotype imputation, overcoming privacy issues with traditional methods. A new two-stage approach using RNN-IMP improves accuracy, even with missing genotype data.
Area of Science:
- Genetics
- Bioinformatics
- Computational Biology
Background:
- Traditional genotype imputation relies on the Li and Stephens model, requiring identifiable reference panels.
- Privacy and consent concerns limit the use of public reference panels.
- Deep learning presents an opportunity for reference-free genotype imputation methods.
Purpose of the Study:
- Introduce deep learning-based, reference-free genotype imputation methods.
- Evaluate the performance of RNN-IMP against existing methods.
- Address the challenge of missing genotype data in imputation.
Main Methods:
- Developed RNN-IMP, a deep learning-based genotype imputation method.
- Evaluated RNN-IMP using the 1000 Genomes Project Phase 3 dataset and simulated SNP data.
- Implemented a two-stage imputation strategy using denoising autoencoders for missing genotypes.
Main Results:
- RNN-IMP demonstrates comparable imputation accuracy to Li and Stephens model-based methods.
- Missing genotype values degrade RNN-IMP's imputation accuracy.
- The two-stage imputation strategy effectively restores accuracy lost due to missing data.
Conclusions:
- Deep learning enables reference-free genotype imputation, mitigating privacy concerns.
- RNN-IMP offers a viable alternative to traditional methods, especially with the proposed two-stage imputation strategy.
- The enhanced RNN-IMP method improves the practical utility of genotype imputation.

