💻 bioinformatics

Reconstructing donor genotypes from scRNA-seq for downstream eQTL and HLA-TCR analysis

The paper introduces scTAPAS, a bioinformatic framework that reconstructs donor genotypes directly from single-cell RNA sequencing reads, enabling accurate downstream eQTL and HLA-TCR association analyses even in the absence of matched conventional genotyping data.

Original authors: Kockelbergh, H., Astley, J., Kichula, K. M., Zhang, Z., Ng, E. S., COvid-19 Multi-omics Blood ATlas (COMBAT) Consortium,, Dendrou, C. A., Chong, A. Y., Watson, R. A., Fairfax, B. P., Knight, J. C., Ba

Published 2026-10-05

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Kockelbergh, H., Astley, J., Kichula, K. M., Zhang, Z., Ng, E. S., COvid-19 Multi-omics Blood ATlas (COMBAT) Consortium,, Dendrou, C. A., Chong, A. Y., Watson, R. A., Fairfax, B. P., Knight, J. C., Band, G., Norman, P. J., Carrington, M. N., Sansom, S. N., Mentzer, A. J., Jostins-Dean, L., Luo, Y.

Original paper dedicated to the public domain under CC0 1.0 (https://creativecommons.org/publicdomain/zero/1.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Inside every drop of blood, millions of tiny cells are busy performing their specific jobs, from fighting infections to carrying oxygen. While scientists have long known that our genes influence how these cells behave, understanding exactly which genes do what has been difficult because cells of the same type can act differently depending on their genetic makeup. For years, researchers could only study these genetic effects by looking at a large, mixed crowd of cells at once, which often hid the unique behaviors of individual cell types. The invention of single-cell sequencing allowed scientists to look at each cell individually, revealing a much more detailed picture of how our biology works. However, this powerful tool hit a wall: to see how genes control cell behavior, researchers usually need a separate, expensive test to read a person's DNA code. Without this separate DNA test, the single-cell data remains a puzzle with missing pieces, limiting the ability to connect genetic differences to cellular functions.

A team of researchers has now found a way to solve this missing piece problem by using the single-cell data itself to reconstruct the donor's genetic code. They developed a method called scTAPAS, which acts like a detective that can piece together a person's genetic identity just by reading the messages the cells are sending. When a cell is active, it reads its DNA and creates RNA, a molecular copy that carries instructions for making proteins. Even though this RNA is a partial and sometimes messy copy of the full genetic code, it still contains enough unique genetic markers to identify the person it came from. The researchers realized that by gathering all the RNA messages from a single person's cells and comparing them against a large library of known human genetic patterns, they could fill in the gaps and accurately predict the person's full genetic profile. This approach means that scientists can now perform genetic studies on existing collections of single-cell data, even if the original researchers never collected a separate DNA sample.

To test if this idea worked in practice, the team applied their method to a massive dataset of blood cells collected from people during the COVID-19 pandemic. They focused on a specific region of the genetic code known to be crucial for the immune system, which is also the most complex and difficult part of the genome to read. By aggregating the RNA reads from each donor and using a statistical tool to infer the missing genetic information, they were able to generate a complete set of genetic predictions for each person. When they compared these predictions to the actual DNA tests that had been done on the same people, the match was remarkably strong. Although the method only looked at a small fraction of the total genetic variants available from standard DNA tests, it successfully recovered nearly seventy percent of the important genetic links between genes and cell behavior that were found using the traditional, more expensive methods. The predictions were so accurate that they could also be used to determine a person's ancestral background, a key step in ensuring that genetic studies are not skewed by differences in population history.

The researchers then pushed the method further to see if it could handle the immune system's most specialized genetic features, known as HLA types. These are highly variable genes that teach the immune system to recognize foreign invaders. Because these genes are so complex and vary wildly between people, they are often missing from standard genetic studies, yet they are essential for understanding how the immune system interacts with viruses and vaccines. Using their reconstructed genetic data, the team was able to predict these HLA types with a high degree of accuracy. They then combined this information with data on the immune cells' receptors, which are the tools cells use to recognize threats. This allowed them to discover specific connections between a person's HLA type and the specific receptors their immune cells were using. For example, they found that certain HLA types were strongly linked to how CD4+ T cells, a type of white blood cell, arranged their receptors. These connections were only visible because the researchers could now see both the genetic code and the cell's behavior in the same dataset.

The study demonstrates that the genetic information hidden inside single-cell RNA data is sufficient to unlock powerful new analyses that were previously impossible without separate DNA testing. While the method works best when there is a large number of cells to analyze and is most accurate for common genetic variants rather than rare ones, it opens the door for re-examining thousands of existing single-cell studies. Researchers can now go back to these datasets and ask questions about how genetics shapes cell behavior and immune responses without needing to collect new samples. This approach does not replace the need for high-quality DNA tests in all situations, but it provides a practical and effective alternative for studies where genetic data was never collected, turning a limitation of past research into a new opportunity for discovery. By extracting the genetic story from the cellular messages themselves, scientists can now explore the intricate relationship between our genes and our immune system with a clarity that was previously out of reach.

Technical Summary: Reconstructing Donor Genotypes from scRNA-seq for Downstream eQTL and HLA–TCR Analysis

Problem Statement
Single-cell RNA sequencing (scRNA-seq) enables the identification of expression quantitative trait loci (eQTLs) within specific cell types, offering insights into genetic regulation obscured in bulk analyses. However, the broader application of single-cell eQTL (sc-eQTL) studies is constrained by the frequent absence of matched genotype data for scRNA-seq cohorts. While genetic variants can be detected directly from scRNA-seq reads, the resulting information is sparse and unevenly distributed due to allelic dropout and expression bias. Consequently, direct variant calling from scRNA-seq is generally insufficient to support robust genetic association analyses or HLA typing, limiting the ability to study cell type-specific regulatory effects and immunogenetic interactions in existing datasets.

Methodology: scTAPAS Framework
The authors present scTAPAS (single-cell Transcriptome Allelic Prediction and Association Studies), a bioinformatic framework designed to reconstruct donor genotypes directly from scRNA-seq reads. The workflow proceeds as follows:

Read Aggregation: Sequencing reads aligned to chromosome 6 are consolidated into donor-specific alignment files, aggregating sparse genetic information across all cells within a donor.
Reference-Based Imputation: The aggregated reads are processed using QUILT2, a read-aware imputation method that utilizes sparse low-coverage sequencing reads and large haplotype reference panels (specifically the 1000 Genomes Project phase 3) to infer genome-wide genotype dosages.
Quality Control (QC): To ensure accuracy, inferred variants undergo strict filtering:
- Imputation quality score (INFO_SCORE) > 0.8.
- Minor Allele Frequency (MAF) > 5%.
- Hardy-Weinberg Equilibrium (HWE) p-value > 1×10⁻⁶.
Downstream Extensions:
- eQTL Mapping: Quality-controlled variants are used for cell type-resolved cis-eQTL mapping using pseudobulk expression profiles.
- HLA Imputation: Variants within the Major Histocompatibility Complex (MHC) region are used to impute classical HLA alleles and amino acid polymorphisms using the Michigan Imputation Server (multi-ethnic HLA v2 panel) and EAGLE2 phasing.
- HLA-TCR Analysis: Imputed HLA types are integrated with matched single-cell T-cell receptor (TCR) sequencing data to test associations between HLA variation and TCR repertoire features.

Key Results
The framework was benchmarked using the COVID-19 Multi-omic Blood Atlas (COMBAT) dataset, comprising 140 samples from 124 donors with matched array-based genotyping, HLA sequencing, and single-cell TCR sequencing.

Genotype Concordance: scTAPAS-derived genotype dosages showed strong concordance with reference datasets. After applying the INFO_SCORE > 0.8 and MAF > 5% filters, the median Pearson $R^2$ between scTAPAS and array-based genotypes was 0.797 (Global Screening Array) and 0.799 (TOPMed-imputed). Concordance was positively correlated with the number of cells per sample and the transcriptional context of the variants.
eQTL Discovery: Despite testing only 18.0% of the variants available from conventional array-based genotyping, scTAPAS recovered 68.6% of the cell type-eGene pairs identified using array-based genotypes. For shared cell type-eGene pairs, there was strong concordance in effect sizes ( $R^2 = 0.936$ ). Furthermore, scTAPAS-only eGenes showed high overlap (80.0%) with findings from the independent OneK1K study, suggesting the recovered signals are reproducible.
HLA Inference: scTAPAS successfully imputed classical HLA alleles. At two-field resolution, it showed modest correlation with gold-standard HLA sequencing ( $R^2 = 0.586$ ), improving to $R^2 = 0.664$ for common alleles. Notably, scTAPAS outperformed the direct transcript-based method arcasHLA for HLA Class I genes (median $R^2 = 0.659$ vs. 0.317), while arcasHLA performed better for Class II genes.
HLA-TCR Associations: Using scTAPAS-derived HLA types, the study identified significant associations between HLA Class II alleles and TCR V gene segment usage in CD4+ T cells (e.g., HLA-DRB1*03 with TRAV12-2). These associations were consistent with those derived from conventional genotype-based HLA imputation. No significant associations were detected in CD8+ T cells, likely due to lower cell counts in the dataset.

Significance and Claims
The paper claims that scTAPAS provides a practical workflow to extend genetic and immunogenetic analyses to scRNA-seq cohorts lacking matched genotyping or HLA typing. By reconstructing donor genotypes from transcriptomic data, the method:

Expands Discovery Potential: It enables cell type-resolved eQTL mapping in datasets where conventional genotyping is unavailable, recovering a substantial proportion of regulatory architecture despite reduced variant density.
Enables Immunogenetic Analysis: It allows for HLA-aware analyses, such as linking HLA variation to TCR repertoires at single-cell resolution, which is critical for understanding immune responses and T-cell development.
Complements Existing Tools: The contrasting performance of scTAPAS and arcasHLA across HLA classes suggests they capture complementary information, with scTAPAS offering a robust alternative for Class I inference in 5' scRNA-seq data.

The authors emphasize that while the method is powerful, it is subject to limitations regarding rare variant detection, dependence on reference panel diversity, and the requirement for sufficient cell numbers and sequencing depth. They conclude that scTAPAS offers a viable route to leverage existing single-cell cohorts for genetic regulation and immunogenetic studies, provided appropriate consent and governance for genetic analysis are in place.

More like this