Mining transcriptomic data for gestational diabetes mellitus: What public datasets reveal

Abstract

Gestational diabetes mellitus (GDM) affects up to 14% of pregnancies globally and remains a major threat to maternal and fetal health, contributing to complications such as preeclampsia, macrosomia, and long-term metabolic disorders. As GDM involves dynamic immune-metabolic changes at the maternal-fetal interface, characterizing its gene expression landscape across relevant tissues is essential to identify biomarkers, therapeutic targets, and disease mechanisms. Over the past several years, numerous transcriptomic studies and publicly available datasets have been generated to uncover molecular mechanisms of GDM across diverse tissues and experimental models. Despite their potential, these resources remain underutilized, with many datasets yet to be fully mined for their biological and clinical insights. This review provides an overview of methodological advances in gene expression profiling relevant to GDM and highlights key tissues and models used. We discuss limitations of current datasets such as the scarcity of single-cell data and metadata inconsistencies and propose future directions, including integrative and cell-type-resolved transcriptomic studies. By consolidating and contextualizing current resources, this article underscores the importance of deeper data reuse and provides a concise roadmap for future research on the pathogenesis of GDM.

Key Words: Gestational diabetes mellitus; Gene expression; RNA-sequencing; DNA microarray; Molecular mechanism

Core Tip: Gestational diabetes mellitus (GDM) affects up to 14% of pregnancies and poses significant risks to both mothers and offspring. This mini-review summarizes publicly available transcriptomic datasets related to GDM across tissues, data types, and platforms, highlighting critical gaps such as the underrepresentation of pancreatic islets and adipose tissue. It also outlines opportunities for integrative, multi-omics, and cell-type resolved approaches. By consolidating current resources, this review provides a roadmap for advancing biomarker discovery, therapeutic development, and mechanistic understanding of GDM pathogenesis.

INTRODUCTION

Gestational diabetes mellitus (GDM) is a metabolic disorder defined by glucose intolerance with onset or first recognition during pregnancy[1-3]. Consistent with international guidelines, GDM is commonly diagnosed using the International Association of Diabetes and Pregnancy Study Groups/World Health Organization 75-g oral glucose tolerance test thresholds fasting plasma glucose ≥ 5.1 mmol/L, 1-hour ≥ 10.0 mmol/L, or 2-hour ≥ 8.5 mmol/L[4-6]. Affecting 14% of pregnancies globally[7], GDM is associated with serious short- and long-term complications for both mother and child[8], including preeclampsia[9], macrosomia[10] and an increased risk of metabolic syndrome later in life[11]. Understanding the underlying molecular mechanisms is therefore of great clinical and public health significance.

Over the past decade, transcriptomic profiling techniques such as DNA microarrays, bulk RNA sequencing (bulk RNA-seq), and more recently, single-cell and epitranscriptomic approaches have enabled comprehensive interrogation of gene expression changes in GDM[12]. However, public transcriptomic data on GDM are fragmented, lack standardization and often do not cover key biological tissues such as pancreatic islets or adipose tissue[13]. While several studies focus on placental dysfunction or circulating biomarkers, a comprehensive summary of available datasets and their research scope is lacking[13,14].

To address this gap, we identified and analyzed gene expression datasets related to GDM. We analyzed dataset characteristics, including organism, tissue source, data type, platform, and the presence of control groups, and classified them according to research themes. This mapping not only provides researchers with a practical reference for selecting data suitable for secondary analyses or cross-cohort comparisons, but also highlights tissues, data types, and methodological approaches that remain underrepresented, thereby drawing attention to research gaps with high potential to advance understanding of GDM pathophysiology. Building on these observations, we also outline opportunities for integrative, multi-omics, and cell type resolved analyses that can maximize the value of current resources and help guide future data-generation strategies. Collectively, these contributions transform a scattered and underutilized body of data into a coherent framework that can accelerate biomarker discovery, therapeutic target identification, and mechanistic insight into GDM.

DATA SOURCES AND LITERATURE SELECTION STRATEGY

Data sources and search strategy

As this article is a narrative minireview, we aimed to summarize representative and single-cell studies related to GDM, rather than to perform a formal systematic review. To ensure clarity in how the literature and datasets were identified, we briefly outline our search approach here.

Relevant publications and datasets were retrieved from PubMed, Gene Expression Omnibus (GEO), ArrayExpress, and Google Scholar. Search terms included various combinations of “gestational diabetes”, “GDM”, “placenta” or “islet” AND “single-cell RNA-seq” or “transcriptome”.

Datasets meeting the following inclusion criteria were considered: (1) Studies focusing on GDM; (2) Publication date between January 2005 and September 2025; (3) English-language publications with publicly accessible raw or processed data; and (4) Transcriptomic or single-cell RNA sequencing (scRNA-seq) datasets with clearly annotated tissue or cell type, species, and sufficient accompanying metadata. When available, ethical approval statements were cross-checked in the original publications.

Additionally, β-cell focused datasets were included for their critical mechanistic insights into β-cell adaptation, proliferation, and dysfunction during pregnancy or metabolic stress, even if not exclusively modeling GDM. The inclusion of these datasets is grounded in the established physiological paradigm that β-cell plasticity is central to gestational glucose homeostasis. Classic work by Kim et al[15] demonstrated that serotonin signaling downstream of placental lactogens drives β-cell proliferation during pregnancy, while more recent single-cell analyses by Chung et al[16] have revealed cell type specific transcriptional adaptations of pancreatic islets across pregnancy and postpartum. Thus, these datasets serve as essential references for understanding β-cell failure in GDM, directly informing the cellular and molecular context of gestational glucose regulation.

Dataset verification and harmonization procedures

All datasets were subjected to a multi-step verification workflow that included: (1) Confirmation of dataset identity by cross-checking GEO/ArrayExpress entries with original publications; (2) Harmonization of sample annotations to ensure consistent naming of tissues, cell types, species, and experimental conditions; (3) Standardized categorization of datasets by organism (human vs mouse) and by tissue type (e.g., placenta, islet, peripheral blood); and (4) Removal of duplicate datasets or re-analyses derived from the same primary study. These procedures ensured accuracy, comparability, and transparency across all included datasets.

BIOLOGICAL SOURCES OF GDM TRANSCRIPTOMES: TISSUE TYPES, SPECIES AND DATASET SCALE

To better understand the biological focus of transcriptomic research in GDM, we surveyed the distribution of sample types, organisms, and corresponding sample sizes across publicly available datasets, as shown in Figures 1 and 2, Table 1 and Supplementary Table 1.

Figure 1 Species and tissues coverage in gestational diabetes mellitus transcriptomic studies. A: Rattus norvegicus. Datasets most frequently sample pancreatic islets, cardiac tissue and placenta; B: Mus musculus. Commonly profiled tissues include hippocampus, pancreatic islets, liver, skeletal muscle and uterus; C: Ovis aries. Sheep studies primarily focus on pancreatic islets; D: Homo sapiens. Human datasets encompass placenta, maternal peripheral blood, chorionic villi, amniocytes and fetal umbilical cord blood (e.g., venous blood).

Figure 2 The dataset counts and sample sizes across four major categories in gestational diabetes mellitus transcriptomic studies. From the innermost to the outermost ring, the concentric layers represent: (1) Broad tissue categories (placenta-related, blood and its components, pancreas/islet-related and others); (2) The number of datasets assigned to each subcategory; and (3) The cumulative sample size for each category. For example, within the placenta-related category, 19 datasets focusing on placenta were identified, together comprising 1189 samples. PBMCs: Peripheral blood mononuclear cells; sEVs: Secreted extracellular vesicles.

Table 1 Overview of transcriptomic datasets included in this mini-review on gestational diabetes mellitus.

Series	Organisms	Tissues	Sample source (M/O/B)	Sequencing type	Research theme	Sample count
GSE267340	Homo sapiens	Placenta	M	scRNA-seq	Placenta-related	4
GSE173193	Homo sapiens	Placenta	M	scRNA-seq	Placenta-related	8
GSE249311	Homo sapiens	Placenta	M	RNA-seq	Placenta-related	54
GSE154414	Homo sapiens	Placenta	M	RNA-seq	Placenta-related	8
GSE144276	Rattus norvegicus	Placenta	M	RNA-seq	Placenta-related	6
E-MTAB-11439	Mus musculus	Liver	M	RNA-seq	Placenta-related	8
E-MTAB-9203	Homo sapiens	Placenta	M	RNA-seq	Placenta-related	14
GSE206042	Homo sapiens	Placenta	M	ncRNA-seq	Placenta-related	6
GSE213799	Homo sapiens	Placenta	O	ncRNA-seq	Placenta-related	14
GSE112168	Homo sapiens	Chorionic villi	M	ncRNA-seq	Placenta-related	12
GSE154415	Homo sapiens	Placenta	M	Multi-omics	Placenta-related	24
GSE206041	Homo sapiens	Placenta	M	ncRNA-seq	Placenta-related	6
GSE2956	Homo sapiens	Placenta	M	Microarray	Placenta-related	1
GSE19649	Homo sapiens	Blood, placenta	M	Microarray	Placenta-related	5
GSE103552	Homo sapiens	Placenta	O	Microarray	Placenta-related	37
GSE51546	Homo sapiens	Umbilical cord	O	Microarray	Placenta-related	12
GSE154413	Homo sapiens	Placenta	M	ncRNA-seq	Placenta-related	16
GSE89497	Homo sapiens	Placenta	M	RNA-seq	Placenta-related	764
GSE200983	Homo sapiens	Placenta	M	RNA-seq	Placenta-related	6
GSE49524	Homo sapiens	Umbilical cord blood	O	Microarray	Placenta-related	6
GSE250374	Homo sapiens	Placenta	M	ncRNA-seq	Placenta-related	12
GSE236335	Mus musculus	Uterus, vagina	M	RNA-seq	Placenta-related	16
GSE87295	Homo sapiens	Umbilical cord	O	Microarray	Placenta-related	10
GSE150621	Homo sapiens	Amniocytes	O	RNA-seq	Circulating biomarkers	14
GSE70494	Homo sapiens	Placenta	M	Multi-omics	Circulating biomarkers	145
GSE70493	Homo sapiens	Placenta	M	Microarray	Circulating biomarkers	63
GSE267259	Homo sapiens	Serum	M	RNA-seq	Circulating biomarkers	8
GSE284329	Homo sapiens	PBMCs	M	RNA-seq	Circulating biomarkers	17
GSE212309	Homo sapiens	Umbilical cord blood	O	scRNA-seq	Circulating biomarkers	7
GSE228990	Homo sapiens	Umbilical cord blood	O	RNA-seq	Circulating biomarkers	60
GSE192813	Homo sapiens	Plasma sEVs	M	ncRNA-seq	Circulating biomarkers	24
GSE216997	Homo sapiens	Plasma	M	ncRNA-seq	Circulating biomarkers	328
GSE203346	Homo sapiens	Umbilical cord blood	B	RNA-seq	Circulating biomarkers	84
GSE243374	Homo sapiens	Serum maternal exosomes, serum placental exosomes	M	Microarray	Circulating biomarkers	28
GSE154377	Homo sapiens	Plasma	M	RNA-seq	Circulating biomarkers	134
GSE92772	Homo sapiens	Blood	M	Multi-omics	Circulating biomarkers	32
GSE98043	Homo sapiens	Plasma	M	Microarray	Circulating biomarkers	4
GSE65737	Homo sapiens	Umbilical cord vein blood	O	Microarray	Circulating biomarkers	6
E-MEXP-3382	Homo sapiens	Blood	M	Microarray	Circulating biomarkers	6
E-MEXP-3966	Homo sapiens	Blood	M	Microarray	Circulating biomarkers	3
E-MEXP-3349	Homo sapiens	Blood	M	Microarray	Circulating biomarkers	17
GSE241770	Mus musculus	Pancreatic islets	M	scRNA-seq	β-cell/islet function	7899
GSE278861	Mus musculus	Pancreatic islets	O	scRNA-seq	β-cell/islet function	9
GSE234741	Mus musculus	Pancreatic islets	M	scRNA-seq	β-cell/islet function	2
GSE289077	Mus musculus	Pancreatic islets	M	RNA-seq	β-cell/islet function	15
GSE237149	Mus musculus	Liver	B	RNA-seq	β-cell/islet function	40
GSE234740	Mus musculus	Pancreatic islets	M	RNA-seq	β-cell/islet function	11
GSE116663	Rattus norvegicus	Pancreatic islets	O	RNA-seq	β-cell/islet function	12
GSE118323	Rattus norvegicus	Pancreatic islets	O	RNA-seq	β-cell/islet function	6
GSE104017	Rattus norvegicus	Pancreatic islets	O	RNA-seq	β-cell/islet function	12
GSE90022	Ovis aries	Pancreatic islets	O	RNA-seq	β-cell/islet function	8
GSE21860	Mus musculus	Pancreatic islets	M	RNA-seq	β-cell/islet function	6
GSE130997	Mus musculus	Pancreatic islets	M	RNA-seq	β-cell/islet function	6
GSE36067	Rattus norvegicus	Pancreatic islets	M	Microarray	β-cell/islet function	14
GSE100645	Mus musculus	Pancreatic islets	M	Microarray	β-cell/islet function	6
GSE255246	Mus musculus	Skeletal muscle	O	RNA-Seq	β-cell/islet function	10
GSE136737	Rattus norvegicus	Cardiac	O	RNA-seq	β-cell/islet function	12
GSE147039	Mus musculus	Hippocampus	O	Microarray	β-cell/islet function	15
GSE194119	Homo sapiens	Blood exosomes	M	Multi-omics	Predictive models	6
GSE182737	Homo sapiens	PBMCs	M	Microarray	Predictive models	12
GSE114860	Homo sapiens	Exosomes	M	ncRNA-seq	Predictive models	28

M: Maternal; O: Offspring; B: Both; PBMCs: Peripheral blood mononuclear cells; sEVs: Secreted extracellular vesicles; RNA-seq: RNA-sequencing; scRNA-seq: Single-cell RNA-sequencing; ncRNA-seq: Non-coding RNA-sequencing.

Tissues and biological compartments

Understanding the biological compartments represented in transcriptomic datasets is essential for interpreting disease-associated molecular signatures in GDM. The available datasets span a broad range of tissues including placenta, blood and circulating components, pancreatic islets, and multiple metabolic organs each offering a distinct window into the pathophysiology of GDM. However, their distribution is highly uneven, with certain tissues heavily profiled and others critically underrepresented. The subsections below examine these tissue sources in detail and highlight their relevance, limitations, and implications for understanding GDM biology.

Placenta-related datasets: Among the surveyed datasets, the placenta emerged as the predominant focus, with more than one-third devoted to this tissue. This predominance underscores its pivotal role at the maternal-fetal interface, where it regulates nutrient exchange, immune tolerance, and endocrine signaling all of which are disrupted in GDM[17,18]. Placental tissue is also easily accessible at delivery, further contributing to its prominence in transcriptomic studies[19,20].

However, placenta-focused datasets capture only a portion of the biology complexity underlying GDM. The placenta is highly heterogeneous, with extravillous trophoblasts, syncytiotropho blasts, villous cytotrophoblasts, endothelial cells, and immune populations exhibiting distinct and sometimes divergent responses to metabolic stress[21]. GDM affects these cell types differently, meaning bulk placental profiles may mask subtype-specific alterations[14]. Moreover, placental transcriptional programs vary markedly with gestational age, and most datasets derive from term samples, limiting insight into early pathogenic events. Maternal glycemia has also shown weak correlations with placental transcriptomic changes in several cohorts, indicating that placental tissue does not fully reflect systemic metabolic dysfunction such as β-cell failure or hepatic insulin resistance[22]. These limitations highlight the need to complement placenta-based studies with data from metabolic tissues and longitudinal sampling.

Blood and circulation-related datasets: Other commonly utilized tissues included peripheral blood, plasma, and cord blood derived endothelial colony-forming cells. These accessible, fluid-based samples provide minimally invasive access to systemic and fetal circulation related transcriptomic changes, making them especially attractive for biomarker discovery[13]. Notably, datasets categorized under blood and its components showed an average sample size of 38, reflecting the growing emphasis on scalable, high-throughput, non-invasive transcriptomic profiling[23,24], particularly through circulating microRNAs (miRNAs) and exosomal RNA[25].

Pancreatic islet-related datasets: There exist thirteen datasets related to pancreatic islet, including 8006 samples, largely derived from rodent models. These datasets enable exploration of β-cell proliferation, insulin secretion, mitochondrial adaptation, and endocrine stress responses under gestational metabolic load. Despite their biological relevance, pancreas-related datasets remain fewer in number than placenta-related ones, indicating a gap between clinical disease mechanisms (β-cell dysfunction) and publicly available transcriptomic resources. The details on these datasets are summarized in Table 1 and Supplementary Table 1.

Other tissues-focused datasets: In contrast, other solid metabolic tissues such as, cardiac, skeletal muscle and liver were underrepresented, despite their direct involvement in glucose metabolism and insulin resistance[26]. Moreover, no datasets were available for adipose tissue, despite its established role in insulin sensitivity, lipid storage and inflammation during pregnancy. Their scarcity represents a critical gap in the current dataset landscape and limits mechanistic insight into the systemic nature of GDM.

Maternal and offspring datasets: In addition, for clarity, animal-based datasets were annotated according to the biological source as maternal, offspring, or both in Table 1.

In reviewing the included datasets, we observed that most human studies focused exclusively on the maternal transcriptome, consistent with the clinical emphasis on placental tissue, maternal blood, and pregnancy-associated biomarkers. In contrast, several animal-based studies particularly those involving mouse models of intrauterine hyperglycemia or gestational diabetes profiled fetal or offspring tissues rather than maternal samples. This distinction is biologically meaningful. Maternal datasets capture pregnancy-associated metabolic dysregulation[27], immune activation[28], and alterations in placental function[29] particularly in pathways regulating nutrient transport, angiogenesis and immune modulation thereby reflecting the immediate physiological disruptions characteristic of GDM. Offspring datasets, by contrast, illuminate fetal developmental programming in response to intrauterine hyperglycemia, including changes in β-cell maturation[30], skeletal muscle metabolism[31], mitochondrial function[32] and transcriptional signatures linked to long-term metabolic vulnerability[33]. Together, these complementary perspectives underline that GDM is not only a maternal metabolic disorder but also a developmental condition with multigenerational impact, reinforcing the need to analyze maternal and offspring compartments in parallel rather than in isolation.

Mechanistic implications of missing metabolic tissues: The scarcity of metabolic tissue datasets has important mechanistic consequences. Pregnancy imposes a substantial metabolic burden that relies heavily on β-cell compensation, in which pancreatic β cells expand their mass and enhance insulin secretory capacity[34,35]. Inadequate β-cell adaptation is a central defect in GDM and precedes overt hyperglycemia, making islet transcriptomic data essential for identifying early molecular drivers of impaired insulin secretion, endoplasmic reticulum stress, mitochondrial dysfunction and ferroptosis susceptibility[36]. Although ferroptosis itself is a biochemical form of regulated cell death, components of ferroptosis-associated stress responses are transcriptionally regulated and can therefore be partially reflected in transcriptomic datasets.

Similarly, hepatic insulin resistance is a key contributor to maternal hyperglycemia. The liver orchestrates gestational changes in gluconeogenesis, lipid mobilization, and ketone metabolism; lack of liver transcriptomes limits our ability to characterize these shifts, understand mother fetus metabolic allocation, and identify dysregulated pathways such as peroxisome proliferator-activated receptor signaling, interleukin-6 mediated inflammation, or oxidative stress[37].

Adipose and skeletal muscle also play foundational roles in maternal metabolic remodeling, including insulin sensitivity, lipid storage, adipokine secretion, and immune infiltration[38]. Without datasets from these tissues, cross-organ integration linking placenta-derived signals to systemic insulin resistance remains speculative.

To address these gaps, rodent models provide a powerful complementary approach. They permit controlled dietary or genetic induction of GDM, longitudinal sampling across gestation, isolation of metabolic tissues at defined time points, and application of single-cell multi-omics[39]. Such models can reveal mechanistic pathways in islets, hepatocytes, myocytes, and adipocytes that are inaccessible in human pregnancy, thereby guiding hypotheses for future human studies.

Species and experimental models

Regarding species distribution, the majority of datasets (67.2%) were derived from Homo sapiens, consistent with the clinical focus and accessibility of human tissues during pregnancy. Human studies primarily focused on placenta, blood, and cord blood, enabling ethically feasible investigation of GDM pathophysiology[17].

In contrast, Mus musculus (21.3%) and Rattus norvegicus (9.8%) were primarily employed in experimental models to study mechanistic aspects of β-cell dysfunction, immune dysregulation, and fetal programming under diabetic conditions. Mouse models, such as high-fat or high-fat/high-sugar diet induced gestational metabolic stress models in C57BL/6 mice, have been used to investigate pregnancy-associated metabolic alterations and β-cell dysfunction in vivo[40], while rat insulinoma cell lines (INS-1/RIN) provide well-established in vitro systems for mechanistic studies of β-cell function and regulation[41]. Rodent models also provided access to ethically inaccessible tissues and developmental time points, underscoring their complementary role in GDM research[42].

Additionally, Ovis aries (1.6%) offers a large-animal model with developmental and placental physiology more similar to humans than rodents[43]. Sheep-based studies primarily focusing on fetal pancreatic islets in the context of intrauterine growth restriction associated gestational metabolic stress provide a valuable intermediate platform for investigating endocrine development and metabolic programming in utero, thus bridging the translational gap between rodent models and human studies[44].

Dataset scale and sample size distribution

Regarding sample size, placental datasets averaged around 62 samples, offering moderate statistical power for differential expression analysis. Blood-based studies were typically smaller (mean approximately 14), limiting interpretability. Conversely, plasma studies showed both high sample sizes and scalability, making them well-suited for biomarker discovery, although less informative for tissue-specific mechanistic exploration[45].

Taken together, current GDM transcriptomic studies emphasize two dominant strategies: (1) Mechanistic dissection of placental dysfunction; and (2) Scalable biomarker discovery using fluid-based transcriptomes.

However, the limited inclusion of key metabolic tissues and small sample sizes in several studies highlight the need for future work to expand greater biological diversity and apply higher-resolution approaches, thereby capturing GDM’s complex, multi-organ etiology.

TRANSCRIPTOMIC DATA TYPES AND PROFILING PLATFORMS

To better characterize the methodological evolution of transcriptomic studies in GDM, we also analyzed the usage of sequencing technologies, and the results are as shown in Figure 3A, Tables 1 and 2.

Figure 3 Various transcriptomic technologies in gestational diabetes mellitus research. A: Overall proportions of transcriptomic technologies represented in gestational diabetes mellitus (GDM) datasets. RNA-sequencing (RNA-seq) remains the most prevalent (43%), followed by DNA microarrays (28%), non-coding RNA-seq (13%), multi-omics (6%) and single-cell RNA-seq (scRNA-seq) (10%). These distributions highlight both the persistence of classical platforms and the growing role of emerging modalities in GDM research; B: Distribution of dataset types across research themes in GDM. Placenta-related datasets dominate, followed by circulating biomarkers and β-cell/islet, while predictive modeling studies remain fewer. RNA-seq (blue) and microarray (light blue) are the most widely used platforms overall, whereas non-coding RNA-seq (light green) and scRNA-seq (dark green) contribute substantially to biomarker and islet-related studies. Multi-omics approaches (pink), though relatively limited, are increasingly adopted in biomarker discovery and predictive modeling. RNA-seq: RNA-sequencing; scRNA-seq: Single-cell RNA-sequencing; ncRNA-seq: Non-coding RNA-sequencing.

Table 2 Overview of RNA sequencing and expression profiling platforms used in gestational diabetes mellitus transcriptomic studies.

Platform	Read configuration/probe architecture	Data output capacity	Advantages	Disadvantages	Frequency	Category
NovaSeq 6000	2 × 50 ~ 2 × 250 bp (2 × 150 bp common)	Ultra-high	High-throughput, good cost/data ratio	Expensive, best for large projects	11	NGS, short-read (Illumina)
HiSeq 4000	2 × 50 ~ 2 × 150 bp	High	Mainstream, reliable	Being replaced by NovaSeq	10	NGS, short-read (Illumina)
HiSeq 2000	2 × 100 bp	Moderate	Stable, classical platform	Older, limited throughput	4	NGS, short-read (Illumina)
HiSeq 2500	2 × 50 ~ 2 × 125 bp (rapid)	Moderate	Flexible (rapid/high modes)	Phasing out	6	NGS, short-read (Illumina)
NextSeq 500	2 × 75 ~ 2 × 150 bp	Middle	Fast run time, flexible	Limited throughput, costlier	4	NGS, short-read (Illumina)
NextSeq 2000	2 × 50 ~ 2 × 150 bp	Middle-high	Improved chemistry, higher quality	Still relatively costly	1	NGS, short-read (Illumina)
Illumina BeadChip	48 k/24 k fixed probes	Ultra-high	Mature, low cost	Lower sensitivity, cannot detect novel transcripts	5	DNA microarray
Affymetrix	3’-biased or whole-transcript probes	High	Mature database support	Lower resolution, narrow dynamic range	4	DNA microarray
Agilent	8 × 60 k format, 20638 genes	High	Cost-effective, widely used	Limited depth of analysis	4	DNA microarray

NGS: Next-generation sequencing.

From microarray to bulk RNA-seq

Early transcriptomic studies of GDM were dominated by microarray platforms [e.g., Gene 1.0 ST, Human Transcriptome Array 2.0 and specialized miRNA or long noncoding RNA (lncRNA) arrays], and Agilent whole-genome microarrays were frequently employed[46], alongside occasional use of NuGO human arrays and circular RNA (circRNA)-specific arrays[47,48] (as shown in Table 2). These platforms provided the earliest genome-wide expression profiles but were constrained by reduced sensitivity and limited dynamic range compared with next-generation sequencing technologies.

Since 2018, a clear methodological transition has occurred, with bulk RNA-seq rapidly becoming the dominant approach. The bulk RNA-seq offers higher sensitivity, broader transcriptome coverage and greater scalability, and it is now widely adopted as the standard high-throughput strategy for transcriptome interrogation in GDM research.

Notably, a temporal gap exists between the publication of the earliest GDM-related studies and the release of corresponding datasets. For example, microarray-based studies were first published in 2003, yet the earliest datasets did not appear until 2005; similarly, RNA-seq articles emerged in 2008, but the first datasets were only deposited in 2010.

Emerging modalities beyond bulk RNA-seq

Beyond conventional bulk messenger RNA-focused RNA-seq, recent years have seen a diversification of transcriptomic methods. Non-coding RNA profiling, particularly miRNA-sequencing (miRNA-seq), has shown steady growth from 2021, driven by increasing interest in circulating miRNAs as biomarkers of GDM[49]. Other modalities such as lncRNA-sequencing and circRNA-sequencing have also been applied, extending the molecular scope of GDM studies.

ScRNA-seq has markedly advanced our understanding of cellular heterogeneity at the maternal fetal interface; however, across-study discrepancies and limited dataset coverage remain evident, and single-cell investigations related to GDM are still emerging[50]. In addition, multi-omics approaches that integrate transcriptomics with DNA methylation or RNA modification profiling have begun to appear, underscoring the trend toward more integrative and system-level analyses. By contrast, epitranscriptomic assays [e.g., methylated RNA immunoprecipitation sequencing (MeRIP-seq)] and other specialized methods remain underrepresented, highlighting untapped opportunities for exploring RNA modifications in pregnancy-related disorders[51].

Here again, discrepancies in timing can be observed: Noncoding RNA (ncRNA)-sequencing (ncRNA-seq) articles first appeared in 2015 but the earliest datasets were not released until 2021, whereas both scRNA-seq and multi-omics studies showed near-synchronous trajectories, with publications and datasets emerging together in 2021 and 2015, respectively. These patterns suggest that earlier modalities often followed a “publication first, dataset later” model, while newer approaches increasingly embrace concurrent data sharing, reflecting broader cultural shifts toward open science and FAIR (findable, accessible, interoperable, reusable) data principles.

To improve accessibility for readers, we have added a concise glossary explaining the major transcriptomic technologies including microarray, bulk RNA-seq, ncRNA-seq, scRNA-seq, MeRIP-seq and spatial transcriptomics, as shown in Supplementary Table 2.

Sequencing platforms and technical implementations

In parallel with evolving data types, the technological platforms employed have also shifted. The Illumina HiSeq series (2000, 2500 and 4000) remains the most frequently represented in GDM datasets (n = 24), reflecting its long-standing role as the mainstream instrument for large-scale transcriptome profiling[52,53]. More recently, the Illumina NovaSeq 6000 (n = 11) has gained prominence owing to its ultra-high throughput and cost-effectiveness, gradually displacing HiSeq in newly generated datasets. NextSeq instruments (e.g., NextSeq 500) were used only occasionally, often for smaller or targeted RNA-seq projects, while a single dataset employed the DNBSEQ-G400. Overall, these trends highlight the interplay between evolving transcriptomic methods and the platforms selected to generate GDM datasets.

Implications for comparative analyses and metadata harmonization

The increasing diversity in transcriptomic methods particularly those capturing small RNAs or RNA modifications reflects the expanding scope of GDM research. At the same time, this methodological heterogeneity underscores the need for harmonized metadata and standardized reporting practices enable robust cross-study comparisons and large-scale integrative analyses[12,54].

RESEARCH THEMES IN GDM TRANSCRIPTOMIC STUDIES

To elucidate the research foci represented by current transcriptomic datasets in GDM, we conducted a title-based thematic analysis and categorized the studies into four major themes, as shown in Figure 3B and Supplementary Table 1. Specifically, datasets were grouped according to tissue origin and study purpose: (1) Placenta-related studies focusing on placental tissue or trophoblast-derived samples; (2) Blood-based circulating biomarker studies using whole blood, plasma, serum, or peripheral immune cells; (3) Pancreatic islet-related datasets derived from pancreatic islets, β-cell models, or other metabolic tissues; and (4) Predictive modeling studies aimed at identifying circulating molecular markers or developing diagnostic or predictive signatures.

The largest category comprised placental-focused studies (23 datasets), highlighting the central role of placental dysfunction in GDM pathophysiology. These investigations examined trophoblast gene regulation[55], vascular remodeling[56] and inflammatory signaling pathways within the placenta[57,58], underscoring the placenta’s pivotal role at the maternal-fetal interface.

Biomarker-oriented studies formed another prominent category (18 datasets), particularly those exploring circulating miRNAs and exosome-derived cargoes from plasma[24], serum[59], or placenta[56] as potential diagnostic tools or mechanistic mediators. This trend reflects a growing interest in developing non-invasive molecular markers for early risk prediction.

In contrast, studies on β-cell and islet biology were comparatively sparse (17 datasets) but provide valuable insights into maternal insulin secretion[60] and fetal endocrine development[44,61]. Although limited in number, these datasets contribute important mechanistic understanding and remain underrepresented relative to placenta- and biomarker-focused research[62].

A small number of studies applied predictive modeling (3 datasets), leveraging transcriptomic features to stratify clinical outcomes (e.g., macrosomia) or to derive diagnostic signatures[63].

Overall, this thematic distribution reveals a current dominance of placenta-focused and biomarker-oriented studies in the GDM transcriptomic field[14]. Meanwhile, mechanistic investigations into pancreatic β-cell function, fetal developmental programming and post-transcriptional regulatory layers remain limited but warrant further exploration in future research[13,22].

PUBLIC TRANSCRIPTOMIC RESOURCES IN GDM: PROGRESS, GAPS AND FUTURE DIRECTIONS

This mini-review offers a focused yet informative overview of transcriptomic datasets related to GDM available in public repositories. Our analysis of 61 datasets reveals that while the field has significantly expanded over the past decades particularly after 2020 there remains substantial heterogeneity in experimental design, tissue sources and annotation quality[64].

A dominant feature of GDM transcriptomic research is the reliance on placenta, which appeared in 23 datasets. This emphasis reflects the placenta’s central role in GDM pathophysiology and its accessibility at delivery[65]. Meanwhile, a smaller but growing number of studies have begun to explore circulating biomarkers (e.g., plasma-derived exosomal miRNAs), leveraging the potential for non-invasive diagnostics[66]. However, key metabolic tissues such as pancreatic islets, skeletal muscle and adipose remain underrepresented in publicly available transcriptomic data, despite their known involvement in GDM’s systemic effects.

From a methodological standpoint, DNA microarray and bulk RNA-seq are the most widely employed technique, but recent studies are beginning to adopt more advanced methods such as miRNA-seq, MeRIP-seq and scRNA-seq[67,68]. Still, single-cell datasets remain scarce and are often limited to small sample sizes or restricted tissue contexts. No spatial transcriptomic datasets were identified, suggesting an opportunity to apply high-resolution spatial methods to investigate placental architecture and fetal maternal interactions in GDM.

In terms of research themes, the field remains focused on biomarker discovery (especially miRNAs and extracellular vesicles) and placental dysfunction. While these are valuable areas, other critical processes such as maternal immune modulation, β-cell compensation and fetal programming of disease risk have received relatively limited transcriptomic exploration. Moreover, epitranscriptomic regulation, which emerged in two recent datasets (e.g., N6-methyladenosine RNA modification), remains an underdeveloped but promising frontier.

To advance beyond current limitations, several technically actionable future directions can be envisioned. First, applying single-cell multi-omics including scRNA-seq, single-cell assay for transposase-accessible chromatin using sequencing, and emerging spatial transcriptomic platforms to placental tissues would enable high-resolution mapping of trophoblast heterogeneity, immune stromal interactions, and functional differences between villous and decidual compartments[69]. Second, longitudinal sampling across trimesters, rather than single time points at delivery, would provide critical insight into the temporal evolution of GDM, allowing identification of early molecular predictors and dynamic immune metabolic adaptations[70]. Third, future studies should integrate transcriptomic data with comprehensive maternal metabolic phenotyping, encompassing insulin sensitivity indices, lipid metabolism, β-cell function and inflammatory biomarkers, to improve mechanistic interpretation. Fourth, targeted single-cell analysis of pancreatic β-cell adaptation during pregnancy represents a particularly promising direction, with potential to clarify pathways leading to insufficient insulin compensation in GDM. Finally, adopting standardized metadata and reporting frameworks such as minimum information about a microarray experiment and minimum information about sequencing experiments will improve dataset interoperability, reduce annotation inconsistency, and facilitate cross-cohort integration within the GDM transcriptomic field.

CONCLUSION

This minireview summarizes key trends, strengths and limitations in the current landscape of publicly available transcriptomic datasets on GDM. While the placental transcriptome and circulating biomarkers have been extensively studied, there is a lack of transcriptomic data covering essential metabolic tissues and advanced methodologies such as single-cell and spatial transcriptomics. To advance the field, we recommend expanding dataset diversity to include maternal metabolic organs (e.g., pancreas, adipose tissue), fetal tissues and longitudinal sampling across gestation. Future studies should adopt cutting-edge technologies such as scRNA-seq, spatial transcriptomics and multi-omics integration (e.g., proteomics, epigenomics) to capture the full biological complexity of GDM. Moreover, standardized metadata reporting including gestational age, diagnostic criteria, treatment exposure and maternal fetal outcomes is essential to ensure comparability and reusability. By integrating machine learning approaches and building harmonized transcriptomic cohorts, researchers will be better positioned to identify reliable biomarkers, stratify GDM subtypes and uncover novel mechanistic insights. Ultimately, this curated resource and synthesis provide a foundation for more integrative, translational and impactful research in the era of precision medicine for GDM.