Allele-specific expression, the phenomenon in which one copy of a gene produces more RNA than the other copy in the same cell, has become one of the most informative readouts in modern genomics. Every person carries two versions of most genes, one inherited from each parent, and researchers once assumed these two copies were expressed more or less equally. That assumption turns out to be wrong surprisingly often. Advances in sequencing technology, computational methods, and single-cell analysis have turned allele-specific expression from a niche observation into a powerful lens for understanding how gene regulation evolves, how diseases develop, and how crops can be improved.
What Allele-Specific Expression Actually Is
In a diploid organism, you have a maternal and a paternal copy of nearly every gene. Allele-specific expression (ASE) occurs when one of those copies is transcribed at a higher rate than the other. The imbalance can range from subtle, with one allele making slightly more RNA, to extreme, where only one allele is active and the other is effectively silent. This imbalance generally arises from differences in the regulatory sequences flanking the gene rather than from the protein-coding sequence itself. A variant in a nearby promoter or enhancer can change how strongly a transcription factor binds, which in turn dials the expression of that particular allele up or down.
Researchers distinguish between “cis” effects, which act on the gene copy sitting on the same chromosome as the regulatory variant, and “trans” effects, which come from factors floating around the cell that can affect both copies. In hybrid organisms where the two parental genomes are well characterized, ASE measured in the offspring directly reflects cis-regulatory divergence, because both alleles share the same cellular environment. A study of natural-population hybrids using RNA sequencing found that about 68% of informative sites showed significant cis-regulatory divergence, while 70% showed trans-regulatory divergence, and over half of all sites showed evidence of both types acting simultaneously.1Genome Biology and Evolution. RNA-Seq Analysis of Allele-Specific Expression, Hybrid Effects, and Regulatory Divergence in Hybrids Compared with Their Parents from Natural Populations The interplay between these forces is a major area of study, because it shapes how organisms adapt and how gene expression differences accumulate between species.
From a Single Binding Site to a Species-Level Difference
One of the most compelling demonstrations of how ASE can trace back to specific regulatory changes comes from fruit fly genetics. In hybrids between Drosophila melanogaster and Drosophila simulans, the gene hunchback shows expression biased toward the melanogaster allele. Researchers narrowed the cause to a single regulatory region near the promoter. In simulans, this region contains an extra binding site for the transcription factor Bicoid, along with a site for Huckebein, which can act as a repressor. When the team used gene editing to swap the regulatory region in melanogaster to resemble the simulans version, the expression bias disappeared, confirming that this one regulatory module was responsible for the allele-specific pattern.2PLoS Genetics. Allele-specific expression elucidates cis-regulatory logic This kind of work shows that ASE is not just a statistical abstraction; it can be pinned to individual DNA changes and functionally validated.
The Reference Bias Problem
Measuring ASE from RNA sequencing data sounds straightforward: count how many reads map to the maternal allele versus the paternal allele at each heterozygous site. In practice, the standard reference genome creates a systematic bias. Reads carrying the reference allele tend to map more successfully than reads carrying the alternate allele, because the reference genome is, by definition, the sequence the aligner is trying to match. This “reference mapping bias” can inflate the apparent expression of one allele and make genuine ASE harder to distinguish from artifact.
Several computational tools have been developed to address this. One approach involves constructing a personalized reference that incorporates both alleles, so neither has a mapping advantage. A tool called Ornaments takes this strategy and adds probabilistic assignment of ambiguously mapping reads, correcting for both reference bias and the related problem of reads that could plausibly come from multiple transcripts or loci.3PubMed Central. Ornaments for efficient allele-specific expression estimation with bias correction A separate pipeline developed for livestock genomics adapted open-license tools into a standardized workflow designed to minimize reference bias while profiling ASE across tissues and cell types, demonstrating robust immune-related ASE patterns in crossbred sheep.4PubMed Central. Elimination of Reference Mapping Bias Reveals Robust Immune Related Allele-Specific Expression in Crossbred Sheep Getting the computational side right matters enormously: a false-positive ASE call can send researchers chasing regulatory variants that do not exist.
What Single Cells Reveal About Allelic Noise
Bulk RNA sequencing averages expression across thousands or millions of cells, which can mask what individual cells are doing. When researchers began looking at ASE in single cells, they found something striking: individual mammalian cells frequently express only one allele of a given autosomal gene, even when the tissue as a whole shows expression from both. A study of mouse preimplantation embryos discovered that 12 to 24% of autosomal genes showed monoallelic expression in any given cell, and this monoallelic pattern appeared random and dynamic, varying considerably even between closely related cells in the same embryo.5PubMed. Single-cell RNA-seq reveals dynamic, random monoallelic gene expression in mammalian cells
This randomness reflects the bursty nature of transcription. Genes are not continuously active; instead, they fire in intermittent bursts, and the two alleles burst independently of each other. A tool called SCALE was developed to model this allele-specific bursting from single-cell RNA sequencing, enabling researchers to detect genes where the two alleles differ in burst frequency or burst size and to identify genes whose alleles burst non-independently.6PubMed Central. SCALE: modeling allele-specific gene expression by single-cell RNA sequencing The distinction matters because non-independent bursting could indicate some form of cross-talk between the two alleles, which would have different biological implications than simple stochastic noise.
Long Reads and Allele-Specific Isoforms
Standard short-read sequencing can tell you that a particular variant position shows allelic imbalance, but it often cannot tell you which full-length transcript the imbalanced reads belong to. Long-read sequencing platforms, which produce reads spanning entire transcripts, change the picture. In maize, researchers used long-read sequencing alongside conventional short reads and found that full-length transcripts could deliver accurate haplotype information with high specificity. One gene provided a vivid example: it produced two isoforms that differed only in how the last exon was spliced, and each isoform was exclusively expressed from a different parental allele. The maternal allele produced one splice form; the paternal allele produced the other.7PubMed Central. Variant phasing and haplotypic expression from long-read sequencing in maize This kind of allele-specific isoform usage is invisible to short-read methods and expands the definition of ASE beyond simple expression-level differences to include qualitative differences in the RNA products each allele generates.
Imprinting and X-Inactivation as Classic ASE Phenomena
The most well-known forms of allele-specific expression are genomic imprinting and X-chromosome inactivation. Imprinting affects over 100 genes whose expression depends on which parent they were inherited from. The mechanism relies on DNA methylation marks laid down in the egg or sperm that persist after fertilization. Enzymes called DNMT3A and its cofactor DNMT3L establish these marks in the germline, and the maintenance enzyme DNMT1 preserves them through cell division. Embryos lacking DNMT1 cannot sustain imprinted expression, confirming that methylation is not just correlated with imprinting but required for it.8Portland Press (Essays in Biochemistry). The influence of DNA methylation on monoallelic expression
X-chromosome inactivation silences most genes on one of the two X chromosomes in female mammals, producing organism-wide ASE across hundreds of genes. But the silencing is not absolute. A landmark study examining 683 X-linked genes across human tissues found that up to a third of these genes escape inactivation to some degree, with expression from both the active and inactive X chromosomes. The extent of escape varies between genes, between individuals, and even between tissues.9PubMed Central. Landscape of X chromosome inactivation across human tissues A more recent analysis refined this picture, suggesting that tissue-specific escape from X-inactivation is rarer than initially thought, affecting roughly 12% of X-linked genes examined, and that most X-inactivation patterns are stable across individuals.10eLife. An extended landscape of human X-chromosome inactivation The tension between these estimates reflects genuine measurement challenges and differences in statistical thresholds, and it matters clinically because genes that escape inactivation may contribute to sex differences in disease susceptibility.
Mapping Disease Variants More Precisely
Genome-wide association studies identify stretches of DNA linked to diseases or traits, but these stretches typically contain dozens or hundreds of variants in tight linkage, and pinpointing the actual causal one is difficult. ASE data offer a way to sharpen the picture. The logic is that if a particular variant truly affects gene regulation, individuals who are heterozygous at that variant should show allelic imbalance in the regulated gene’s expression. A method called PLASMA integrates conventional expression-level association data with allele-specific signal, and in application to kidney tumor RNA sequencing, it achieved greater statistical power with 50 samples than standard approaches achieved with 500 samples. PLASMA yielded a nearly sevenfold reduction in the number of candidate causal variants compared to methods that ignore allelic imbalance.11The American Journal of Human Genetics. Allele-Specific Expression: Emerging Trends and Impact in Genomics Similar methods applied to the large Genotype-Tissue Expression (GTEx) dataset achieved an 11% median reduction in the set of putative causal variants.12PLOS Genetics. Leveraging allelic imbalance to refine fine-mapping for eQTL studies
These improvements are not academic bookkeeping. A smaller set of candidate variants means fewer expensive functional experiments needed to validate which variant truly drives a disease association, and it can point researchers toward the right gene and mechanism faster.
Cancer and Immune Evasion
Tumors are riddled with allele-specific events. Chromosomal regions are frequently gained or lost during tumor evolution, creating widespread imbalances between the maternal and paternal copies of the genome. A tool called ASCAT generates genome-wide allele-specific copy number profiles from tumor samples, revealing patterns of gains, losses, and copy-number-neutral events. Applied to breast cancer, ASCAT showed that the basal-like subtype has a significantly higher frequency of loss of heterozygosity (LOH) compared to other subtypes, with large-scale genomic loss followed by whole-genome duplication resulting in near-triploid genomes.13PubMed Central. Allele-specific copy number analysis of tumors
One of the most consequential allele-specific events in cancer involves the HLA genes, which encode proteins that present fragments of abnormal proteins to the immune system. If a tumor loses the copy of an HLA gene that happens to present tumor-specific peptides to immune cells, those immune cells can no longer recognize the tumor. In a study of 90 non-small-cell lung cancers, 40% showed loss of one HLA allele, with the event enriched in the squamous cell subtype, where 61% of tumors were affected. The researchers concluded that HLA loss of heterozygosity may be a more common mechanism of immune evasion than mutations in HLA genes themselves.14Cell. Essential Information about LOHHLA and HLA Loss in NSCLC Understanding which allele is lost has direct implications for immunotherapy, because treatments designed to boost immune recognition depend on the tumor still presenting the relevant peptides.
Tissue-Specific Regulation and Splicing
ASE is not a fixed property of a gene; it can differ across tissues, developmental stages, and environmental conditions. The GTEx project has generated the most comprehensive human atlas to date, containing over 431 million measurements of allelic expression at the variant level across 54 human tissues from over 15,000 samples.15PubMed Central. A vast resource of allelic expression data spanning human tissues Cross-tissue analyses have revealed cases where ASE present in most tissues disappears in a specific one, suggesting tissue-specific regulatory escape.16Bioinformatics. Assessing allele-specific expression across multiple tissues from RNA-seq read data
Beyond expression levels, allele-specific effects extend to alternative splicing. Using GTEx data, researchers identified genetically modulated allele-specific splicing events, where a genetic variant near a splice site causes the two alleles to be spliced differently.17PubMed Central. Allele-specific alternative splicing and its functional genetic variants in human tissues Work in mouse hybrids confirmed that alternative splicing differences between species are predominantly driven by cis-regulatory variants, including variants at and beyond the canonical splice sites.18PubMed Central. Predominant contribution of cis-regulatory divergence in the evolution of mouse alternative splicing This means that a single genetic variant can change not just how much of a gene is expressed from each allele, but which version of the protein each allele produces.
Dynamic ASE in Immune Activation
Some of the most interesting ASE patterns are not static but shift in response to cellular stimulation. A study of T cells found widespread dynamic allele-specific expression across the genome during T cell activation, where the balance between alleles changed over time as the cells responded. Genes showing this dynamic behavior were enriched fourfold within loci previously linked to autoimmune diseases, and pervasive dynamic regulatory effects were found within six HLA genes.19PubMed Central. Allele-specific expression changes dynamically during T cell activation in HLA and other autoimmune loci This finding is important because it means that resting-state measurements of ASE may miss regulatory variants that only manifest during immune responses, which is precisely when autoimmune-relevant gene regulation matters most.
Evolutionary Adaptation Through Regulatory Divergence
ASE measured in hybrids between adapted and non-adapted populations provides a window into how organisms evolve in response to their environment. In maize, F1 hybrids between highland and lowland populations from Mexico and South America were used to identify genes with differential ASE between highland and lowland alleles. The analysis found hundreds of genes with cis-regulatory divergence in each continental population, and 20 of those genes showed convergent cis-regulatory evolution between the two continents, meaning both Mexican and South American highland populations independently evolved expression changes at the same genes.20Molecular Biology and Evolution. Allele-specific Expression Reveals Multiple Paths to Highland Adaptation in Maize
A parallel story emerges in the marine copepod Tigriopus californicus, where populations from different thermal environments were crossed and ASE was measured under control and heat stress conditions. More genes showed allelic imbalance under heat stress than under control conditions, and the increase was specifically driven by overexpression of alleles from the more heat-tolerant population, suggesting that regulatory divergence at these loci is adaptive.21Journal of Heredity. Allele-Specific Expression and Evolution of Gene Regulation Underlying Acute Heat Stress Response and Local Adaptation in the Copepod Tigriopus californicus
ASE and Crop Vigor
Heterosis, the phenomenon in which hybrid offspring outperform their parents, has been exploited in agriculture for over a century, but its genetic basis remains debated. ASE analysis has become a key tool for dissecting what happens in hybrid genomes. In an elite rice hybrid, researchers identified over 3,200 genes with allele-specific expression across multiple tissues and growth conditions. These fell into two categories: “consistent” genes where one parental allele was always favored, and “inconsistent” genes where the bias switched direction depending on the tissue or environment. The consistent genes may produce dominance effects, while the direction-shifting genes could underlie overdominance, where the hybrid state itself is advantageous.22PubMed Central. Patterns of genome-wide allele-specific expression in hybrid rice and the implications on the genetic basis of heterosis
In wheat hybrids, transcriptomic analysis revealed a pronounced expression bias toward alleles from the superior parent, with dominant alleles from that parent accounting for more than 60% of non-additive genes. The bias increased during development from anthesis to grain filling, suggesting a temporal dimension to how hybrid vigor plays out.23PubMed Central. Transcriptome dynamics and allele-specific regulation underlie wheat heterosis at the anthesis and grain-filling stages In maize, allele-specific expression genes were found to be significantly overrepresented among genes previously linked to heterosis-related traits like plant height and ear weight.24PubMed Central. Natural variations of heterosis-related allele-specific expression genes in promoter regions lead to allele-specific expression in maize Collectively, these findings are pushing breeders toward thinking about regulatory variation, not just which protein-coding alleles a crop carries, but how actively each allele is expressed in the hybrid context.
Spatial Transcriptomics and ASE in Tissue Architecture
The newest frontier in ASE research adds a spatial dimension. Technologies like Visium and Slide-seq capture gene expression data while preserving the physical location of each measurement within a tissue section. A computational framework called spASE was developed to detect spatial patterns in allele-specific expression from these data, using a hierarchical model that accounts for cell type mixtures and low signal. Applied to mouse brain tissue, spASE revealed spatial and cell-type-specific allelic imbalance in the cerebellum and hippocampus.25Genome biology. Detection of allele-specific expression in spatial transcriptomics with spASE This matters because regulatory variants could affect expression differently depending on where in a tissue a cell sits, and spatial ASE analysis is the only way to detect that.
Beyond Transcription: Allele-Specific RNA Modifications
ASE research has traditionally focused on how much RNA each allele produces, but recent work extends the concept to chemical modifications on the RNA itself. Using long-read sequencing and machine learning, researchers detected allele-specific patterns of N6-methyladenosine (m6A), the most abundant internal modification on messenger RNA. They found that allelic differences in the sequences surrounding the modification site could alter m6A deposition, as expected, but they also discovered allele-specific modification events where no genetic variant was located near the modified site. This finding would have been impossible with short-read approaches and points to a layer of allele-specific regulation that operates after the RNA is made.26PubMed Central. Long-read RNA sequencing reveals allele-specific N6-methyladenosine modifications
Non-coding RNAs also participate in allele-specific regulation. A study found that a long non-coding RNA called SeT and the microRNA miR-155 jointly regulate the allelic expression profile of the inflammatory gene Tnfα, controlling mRNA dosage through a mechanism involving physical pairing of the two gene copies.27PLoS ONE. Long non-coding RNA SeT and miR-155 regulate the Tnfα gene allelic expression profile These examples suggest that allele-specific regulation is a multi-layered phenomenon that extends well beyond transcription factor binding.
Machine Learning Enters the Picture
Can you predict whether a given genetic variant will cause allele-specific expression without actually measuring RNA? Researchers have shown that machine learning models trained on genomic annotations can predict ASE from DNA sequence features alone, achieving high specificity though with limited sensitivity. Models trained on one dataset and validated on an independent dataset showed comparable performance and relied on similar feature sets, suggesting that the regulatory grammar underlying ASE is reproducible across cohorts.28Scientific Reports. Feasibility of predicting allele specific expression from DNA sequencing using machine learning A newer deep learning approach called DeepAllele takes paired sequences from two alleles at the same locus as input and directly predicts allele-resolved signal intensities and their ratios, learning the subtle combinatorial rules of cis-regulatory grammar.29bioRxiv. Deep genomic models of allele-specific measurements If these models improve in sensitivity, they could eventually allow researchers to prioritize variants for functional testing without generating RNA sequencing data for every tissue and condition of interest, substantially reducing the cost of regulatory variant discovery.
Drug Response and Pharmacogenomics
The clinical relevance of ASE extends into how individuals respond to medications. In type 2 diabetes, researchers investigated whether genetic variants associated with metformin response showed allele-specific expression in drug-naïve patients. Using genotyping and allele-specific expression assays, they found evidence linking genetic variations, perturbed gene expression patterns, and ASE to differences in how patients responded to metformin.30PubMed Central. Implications of genetic variations, differential gene expression, and allele-specific expression on metformin response in drug-naïve type 2 diabetes This kind of work is still early-stage, but it illustrates the potential: if a regulatory variant causes one allele of a drug transporter or metabolic enzyme to be expressed at a lower level, the effective dose of a medication could differ between genotypes in a way that traditional pharmacogenomic testing, which often focuses on protein-coding variants, would miss entirely.