Author: Jihoo Hyun

Affiliation: Canyon Crest Academy, San Diego, California, United States

Acceptance date: 19.07.2026

Publication date: 17.10.2026

Abstract

Human brain organoids are three-dimensional stem-cell-derived models that reproduce selected features of early human neural development while remaining experimentally accessible. Their value has expanded as advances in patterning, tissue architecture, single-cell multi-omics, and genome engineering have made it possible to connect developmental phenotypes with cell-type-specific molecular mechanisms. This review evaluates recent progress in four linked areas: engineering and reproducibility of neural organoid models; single-cell and multi-omic approaches for measuring developmental fidelity; CRISPR-based functional genomics; and applications to neurodevelopmental disorders, with emphasis on autism spectrum disorder and microcephaly. Direct organoid studies are distinguished from primary human brain reference datasets, which are used primarily as benchmarks for developmental fidelity. Recent work demonstrates complementary strengths across unguided cerebral organoids, guided region-specific organoids, single-rosette systems, and assembloids, but also shows that no single model simultaneously maximizes cellular diversity, structural reproducibility, and maturation. Multi-omic studies can reconstruct gene-regulatory programs within organoids, whereas pooled CRISPR screens can test disease-associated genes at scale in developmentally relevant cell states. At the same time, interpretation depends on careful biological replication, avoidance of pseudoreplication, control of culture-related stress, and comparison with primary human tissue. Current brain organoids are established research tools for mechanistic studies and perturbation-based screening, but personalized diagnostics, regenerative applications, and clinical-grade organoid platforms remain prospective. Continued progress therefore requires technical standardization, transparent reporting, rigorous source validation, and proportionate ethical oversight as organoid systems become more complex.

Keywords: brain organoids; neurodevelopmental disorders; single-cell RNA sequencing; multi-omics; CRISPR; autism spectrum disorder; microcephaly

1. Introduction

Direct experimental access to the developing human brain is limited by ethical and practical constraints, and animal models cannot reproduce every feature of human cortical development. Human pluripotent stem cell-derived neural organoids address part of this gap by allowing cells to self-organize or respond to defined patterning cues in three-dimensional culture. Current nomenclature distinguishes organoids by their intended regional identity and assembloids by the controlled combination of multiple region-specific or lineage-specific components [1]. These systems do not reproduce an entire brain; rather, they model selected developmental processes such as progenitor expansion, regional specification, neuronal differentiation, migration, and early network formation.

The field has moved beyond simply asking whether organoids contain recognizable neural cell types. A central question is now how faithfully and reproducibly a model represents the developmental process under study. Cerebral organoids first demonstrated that patient-derived tissue could reproduce aspects of human cortical development and microcephaly in vitro [2]. More recent work has shown that tissue architecture itself influences developmental timing [3], that single-rosette approaches can reduce structural variability [4], and that guided cortical spheroid protocols can achieve substantial consistency across differentiations [5]. These advances make the choice of organoid architecture an experimental variable rather than a purely technical detail.

In parallel, single-cell transcriptomics and chromatin-accessibility assays have enabled direct measurement of developmental trajectories within organoids. Functional genomics has added a complementary layer: CRISPR-based perturbation can test whether a candidate gene or regulatory program causally alters cell fate, migration, or disease-relevant phenotypes. The combination of engineered tissue models, molecular profiling, and perturbation is therefore especially useful for neurodevelopmental disorders (NDDs), in which pathogenic effects can be cell-type-specific and developmentally timed.

This review focuses on human neural organoids as models of neurodevelopment. It compares major model classes, evaluates organoid-focused single-cell and multi-omic studies, summarizes CRISPR-based functional approaches, and examines applications to autism spectrum disorder (ASD) and microcephaly. Primary fetal or postmortem brain datasets are discussed only when they provide an external benchmark for organoid fidelity. Neurodegenerative disorders are outside the disease-specific scope of this review.

2. Methods/Approach to the Review

This article is a focused narrative review rather than a systematic review or meta-analysis. Literature was identified through PubMed, Google Scholar, and publisher databases using combinations of the terms “brain organoid,” “cerebral organoid,” “neural organoid,” “cortical organoid,” “assembloid,” “single-rosette,” “single-cell RNA sequencing,” “multi-omics,” “chromatin accessibility,” “CRISPR,” “autism spectrum disorder,” “microcephaly,” “reproducibility,” and “ethics.” The search emphasized work published from 2015 onward while retaining earlier seminal studies that established major organoid methods. A final targeted literature update was conducted on 12 August 2026.

Studies were prioritized when they met at least one of the following criteria: (1) directly introduced or benchmarked a human neural organoid or assembloid method; (2) used single-cell, multi-omic, or spatial measurements directly in organoids; (3) used genetic perturbation to test developmental mechanisms in organoid or assembloid systems; (4) modeled an NDD using patient-derived or engineered human organoids; or (5) addressed reproducibility, maturation, or ethical oversight relevant to neural organoid research. Primary human brain datasets were included only when they were explicitly used as reference standards for evaluating organoid identity or developmental fidelity. Studies focused only on non-neural organoids, animal-only models, or neurological conditions outside the neurodevelopmental scope were not used as central disease examples.

Peer-reviewed primary research was prioritized for mechanistic claims, with reviews and consensus documents used for terminology, methodological context, reproducibility, and ethics. Preprints were considered during screening but were not used as the sole support for major conclusions when a peer-reviewed version was available. Because this is a narrative review, the search was not designed to yield an exhaustive study count; instead, studies were selected to represent major methodological advances and to directly support the claims discussed.

3. Engineering and Refinement of Brain Organoids

Brain organoid design involves trade-offs among cellular diversity, regional control, structural reproducibility, and experimental complexity. Unguided cerebral organoids permit broad self-organization and can generate multiple neural identities, but this diversity can be accompanied by batch-to-batch and organoid-to-organoid variability. Guided protocols use defined signaling cues to restrict identity toward regions such as dorsal forebrain or ventral forebrain, improving interpretability for region-specific questions. Single-rosette systems further constrain early architecture, whereas assembloids deliberately combine separately patterned components to study interactions that a single regional organoid cannot capture. These model classes are complementary rather than hierarchical (Table 1).

Table 1. Comparison of major human neural organoid model classes.

Comparison	Unguided cerebral organoid	Guided region-specific organoid / spheroid	Single-rosette organoid	Assembloid
Patterning / architecture	Minimal exogenous regional patterning; broad self-organization	Defined cues bias tissue toward a target region	Early organization around one neural rosette / lumen	Separately patterned regions or lineages are fused
Reproducibility	Moderate to variable; sensitive to line and batch effects	Often higher for targeted identity when protocol and QC are standardized	High structural consistency for early architecture in validated protocols	Depends on component consistency; fusion adds another variable
Cellular / regional complexity	Broad diversity; multiple neural identities may coexist	Narrower regional breadth with clearer target identity	Controlled cortical/telencephalic organization with multiple neural lineages	High interaction complexity across selected regions or lineages
Maturation	Mostly fetal-like; prolonged culture can extend maturation	Protocol dependent; supports longitudinal maturation	Progressive neuronal/glial differentiation, still developmentally limited	Can model migration, connectivity, and cell-cell interactions
Key advantages	Broad discovery and human-specific developmental phenomena	Regional interpretability and standardized cohort comparisons	Strong genotype comparisons and morphology-sensitive studies	Direct study of cross-region processes and dynamic phenotypes
Main limitations	Variable identity and architecture can complicate comparison	Reduced diversity may omit cross-region interactions	Constrained architecture may omit multi-region diversity	Fusion timing, geometry, and component quality require control
Representative uses	Microcephaly and broad developmental surveys [2,3]	Cortical development and region-specific disease phenotypes [5,20]	SHANK3 and reproducible telencephalic development [4,16]	Interneuron migration and CRISPR migration screens [6,7]

Note: “Reproducibility” is relative to a defined protocol and does not remove the need for independent cell lines, differentiations, and batches. Model classes are complementary; no row represents a universal quality ranking.

3.1. Morphogenetic fidelity and reproducibility

Neuroepithelial organization is not merely a visual quality-control feature. Chiaradia and colleagues showed that tissue morphology influences the temporal program of human brain organoid development: perturbations in architecture altered spatial organization and developmental progression even when cells shared the same starting genetic background [3]. This finding argues against treating all multi-rosette tissue as intrinsically “disorganized.” Instead, morphology should be evaluated relative to the developmental question and protocol-specific expectations.

Single-rosette approaches attempt to reduce one source of heterogeneity by establishing a more uniform early neuroepithelial unit. Tidball and colleagues described self-organizing single-rosette cortical organoids with a centralized lumen and reproducible early cortical organization [4]. Related single-neural-rosette telencephalic organoids can generate pallial and subpallial progenitors as well as excitatory and inhibitory neuronal lineages in a predictable architecture [16]. These models are particularly attractive when the experiment requires organoid-to-organoid structural consistency or direct genotype comparisons.

Guided cortical spheroid systems provide another route to reproducibility. Yoon and colleagues differentiated multiple pluripotent stem cell lines for more than 100 days and reported high reliability across differentiations using a region-specific cortical spheroid protocol [5]. Reproducibility therefore depends not only on whether a model is “guided” or “unguided,” but also on the quality of starting cell lines, differentiation design, batch structure, and prespecified quality-control measures.

3.2. Region-specific organoids and assembloids

Region-specific models allow investigators to isolate developmental programs that are difficult to interpret in heterogeneous cerebral organoids. A classic example is dorsal-ventral forebrain fusion. Xiang and colleagues generated regionally specified organoids and fused dorsal cortical and ventral medial ganglionic eminence-like tissue to model human interneuron migration [6]. This type of assembloid makes it possible to study a process – tangential migration from ventral to dorsal forebrain – that inherently depends on interactions between distinct regional identities.

The same architecture supports functional screening. Meng and colleagues used a doxycycline-inducible CRISPR-Cas9 system to investigate 425 NDD-associated genes. A generation screen in ventral organoids identified regulators of interneuron production, whereas a migration screen across more than 1,000 dorsal-ventral forebrain assembloids identified genes involved in motility and organelle positioning, including LNPK [7]. This study illustrates an important advantage of assembloids: tissue fusion can convert a complex developmental event into a measurable phenotype while preserving region-specific context.

3.3. Functional maturation

Maturation remains a relative concept in organoid research. Electrophysiological activity does not imply adult-like tissue or organism-level brain function. Trujillo and colleagues showed that cortical organoids can develop increasingly complex oscillatory network activity over time, providing a tractable system for studying early circuit emergence [8]. Long-term culture can extend developmental progression: Gordon and colleagues found that transcriptomic, epigenetic, and physiological features of cortical organoids followed key early postnatal transitions over prolonged culture, with several molecular measures reaching postnatal-like stages after approximately 250-300 days [9]. Even so, many cell types and circuit properties remain immature relative to the postnatal human brain.

Bioengineering strategies can improve viability and cellular support. For example, Cakir and colleagues engineered vascular-like features in human brain organoids, illustrating one approach to mitigating diffusion and metabolic limitations [22]. Such interventions can improve specific aspects of physiological relevance, but they also introduce new sources of technical variability. Consequently, maturation should be reported using explicit molecular, cellular, or electrophysiological criteria rather than as a single global property.

Figure 1. Integrative workflow for neural organoid studies of neurodevelopmental disorders. _{Solid arrows represent work within organoid systems, moving from model architecture to profiling, perturbation, and disease applications. Primary human brain data serve as external developmental benchmarks; Gordon et al. performed downstream CRISPR validation in induced human neural progenitor cells, a separate experimental system [18]. Diagram is conceptual and does not imply that every cited study used all methods.}

4. Single-Cell and Multi-Omic Profiling of Organoids

Single-cell methods are most informative when they are applied directly to organoid tissue and interpreted against appropriate developmental references. Primary human brain datasets remain valuable, but they should be distinguished from evidence generated in vitro. In this review, primary tissue is treated as an external benchmark rather than as direct evidence about organoid behavior.

4.1. Measuring organoid developmental trajectories

Single-cell RNA sequencing can resolve progenitor, neuronal, and glial states within an organoid and can test whether those states follow expected developmental trajectories. It also exposes model-specific artifacts. Bhaduri and colleagues compared cortical organoids with primary developing human cortex and found that organoids ectopically activated cellular stress pathways that impaired molecular subtype specification; transplantation into mouse cortex alleviated some of these stress-associated defects [10]. The study provides a critical caution: an apparent transcriptional difference between an organoid and primary tissue may reflect culture stress rather than a disease mechanism.

A broader quantitative framework is provided by the Human Neural Organoid Cell Atlas. He and colleagues integrated 36 single-cell transcriptomic datasets spanning 26 organoid protocols, totaling more than 1.7 million cells, and mapped organoid states to developing human brain references [12]. Rather than assuming fidelity from morphology alone, this approach estimates which primary cell types and states are reproduced across protocols. The atlas also highlights that different protocols cover different portions of neural developmental space, reinforcing the need to match model selection to the biological question. Wang and colleagues assembled a biobank of 352 patient-derived induced pluripotent stem-cell lines for NDDs and phenotyped more than 6,000 organoids from 35 representative patients and 10 healthy controls by histology and single-cell transcriptomics [26]. The study provides a complementary, patient-based atlas of cellular and tissue phenotypes across disorders; its organoid count must not be mistaken for the number of independent donors.

4.2. Multi-omic reconstruction of gene regulation

Fleck and colleagues directly profiled single-cell transcriptomes and accessible chromatin across a time course of brain organoid development [11]. They developed Pando, a framework that integrates transcription-factor expression, chromatin accessibility, and predicted regulatory elements to infer gene-regulatory networks. The resulting atlas captured regional and temporal changes in regulatory programs during organoid regionalization and neurogenesis. This is a stronger use of multi-omics than simply placing an organoid result next to an unrelated primary-tissue atlas because the molecular modalities are measured in the organoid system itself.

The same study also demonstrates why multi-omics and perturbation are complementary. Regulatory networks inferred from correlated molecular measurements are hypotheses; functional perturbation can then test whether predicted regulators alter cell fate. Similarly, Li and colleagues combined the CHOOSE pooled CRISPR system with single-cell transcriptomics in mosaic cerebral organoids and generated single-cell multiome data to construct developmental regulatory networks [13]. By perturbing 36 high-risk ASD genes involved in transcriptional regulation, they identified cell types and fate transitions that were especially sensitive to genetic disruption. These studies move the field from descriptive atlases toward causal regulatory models. Spatial measurements add information that dissociated-cell profiles cannot retain. Lin and colleagues combined single-cell RNA sequencing with imaging-based spatial transcriptomics in patient-derived ASD cortical organoids and identified abnormal progenitor–neuron zoning and local patches of neuronal disorganization that varied among patients [27]. Their findings show how regional heterogeneity within an organoid can be obscured by an aggregate expression profile.

4.3. Genetic background and inter-individual variation

One limitation of conventional case-control organoid experiments is that each organoid usually represents a single donor background. Brain Chimeroids provide a complementary design in which cells from multiple donors are reaggregated into the same cortical organoid, exposing different genotypes to a shared culture environment. Antón-Bolaños and colleagues used this approach to examine responses to ethanol and valproic acid and observed donor-specific, cell-type-specific differences in susceptibility [14]. Chimeroids do not replace isogenic models; instead, they are useful for separating environmental exposure effects from inter-individual genetic variation within a common experimental context.

Together, direct organoid multi-omics, large integrated atlases, patient-derived phenotypic biobanks, spatial profiling, and multi-donor designs address different questions. Time-course multi-omics identifies regulatory trajectories within a model; integrated atlases benchmark fidelity across protocols; patient-based atlases compare phenotypes across disorders; spatial analysis tests local architecture; and chimeroids test genetic-background effects under shared conditions. Their conclusions should therefore be compared rather than treated as interchangeable.

5. CRISPR-Based Functional Genomics in Organoids

Genome engineering strengthens organoid studies by converting disease-associated variants and regulatory hypotheses into experimentally testable perturbations. However, the interpretation of an edited phenotype depends on the direction and magnitude of the genetic change, the cell type in which it acts, and the developmental stage at which it is measured.

5.1. Targeted disease modeling

Several ASD-associated genes illustrate the importance of gene-specific interpretation. Paulsen and colleagues engineered haploinsufficiency of KMT5B, ARID1B, and CHD8 across multiple human donor backgrounds and found shared but asynchronous abnormalities in the development of GABAergic neurons and deep-layer excitatory projection neurons [15]. This organoid evidence ties CHD8 haploinsufficiency to altered developmental timing.

SHANK3 provides a different example. Wang and colleagues generated telencephalic organoids from single neural rosettes and used control and SHANK3-deficient lines to study telencephalic development [16]. The model showed reproducible cellular composition and revealed molecular and synaptic consequences of SHANK3 deficiency within a human organoid context.

PTEN illustrates why the direction and genetic context of perturbation must be stated precisely. Fu and colleagues used isogenic cortical organoids to study the heterozygous PTEN p.Ile135Leu variant in ASD and control backgrounds; the variant increased organoid size at week 4 in both control and ASD genetic backgrounds. Later growth and neurodevelopmental effects differed by background: at week 8, variant-associated enlargement persisted in the control background but was not observed in the ASD background, whereas increased progenitor and neuronal production was observed in the ASD background but not in the control background [28]. Separately, Dhaliwal and colleagues modeled mild PTEN overexpression and observed reduced neural precursor proliferation, premature neuronal differentiation, reduced AKT activation, and smaller organoids with microcephaly-like features [17]. These are different perturbations and should not be treated as interchangeable.

5.2. High-throughput pooled screens

Pooled CRISPR screening allows many candidate genes to be tested in parallel. Li and colleagues developed CHOOSE – CRISPR-human organoids-single-cell RNA sequencing – for pooled loss-of-function screening in mosaic cerebral organoids [13]. Perturbation of 36 high-risk ASD genes revealed effects on cell-fate determination and highlighted vulnerable populations including dorsal intermediate progenitors, ventral progenitors, and upper-layer excitatory neurons. Because edited and unedited cells coexist in the same organoid, mosaic designs can reduce some culture-level confounding while enabling cell-resolved phenotyping.

Assembloid screens extend this approach to dynamic behavior. Meng and colleagues separated interneuron generation from migration into two related screens and used dorsal-ventral fusion for the migration assay [7]. This distinction is experimentally important: a gene that reduces the number of interneurons is not necessarily a migration gene, and a valid migration screen must measure movement after ventral and dorsal tissues have been brought into an assembloid context.

Microcephaly has also been studied using pooled tissue-level genetic screening. Esk and colleagues performed a CRISPR loss-of-function screen in three-dimensional human neural tissue to identify regulators of brain size and highlighted endoplasmic-reticulum-related mechanisms through IER3IP1 [19]. The study is especially relevant to microcephaly and functional screening because it links candidate genetic perturbations to growth phenotypes directly in a human three-dimensional neural system.

5.3. From network inference to causal validation

Gordon and colleagues profiled cortical organoids derived from 70 quality-controlled hiPSC lines from 55 individuals, encompassing eight ASD-associated mutations, idiopathic ASD, and unaffected controls at four time points through day 100 [18]. The strongest mutation-specific expression changes occurred early; shared transcriptional changes became more prominent during development among genetically defined forms of ASD. The study did not detect a significant shared signal across its idiopathic ASD cases. That result is limited to the cohort and assays studied and does not establish that convergence is absent in idiopathic ASD generally.

Gordon and colleagues then used CRISPR interference in induced human neural progenitor cells to test candidate transcriptional regulators [18]. The organoid cohort comprised patient-derived and unaffected hiPSC lines; engineered perturbations belonged to downstream validation in a different cell system. The study thus connects organoid-based discovery to neural-progenitor validation without implying that the validation screen was conducted in organoids.

6. Applications to Neurodevelopmental Disorders

The strongest disease applications of organoids occur when the model captures a developmental process plausibly linked to the disorder and phenotypes are supported by appropriate controls. ASD studies investigate gene-specific effects and possible shared developmental mechanisms; microcephaly models often permit direct measurement of progenitor behavior and tissue growth.

6.1. Autism spectrum disorder

ASD is genetically heterogeneous, so no single organoid phenotype can represent the entire condition. Instead, organoids are useful for asking whether distinct risk genes converge on common cell types or developmental processes. Paulsen and colleagues showed that KMT5B, ARID1B, and CHD8 haploinsufficiency produced gene-specific timing differences but converged on abnormalities affecting shared neuronal classes [15]. Li and colleagues independently used pooled perturbation of 36 high-risk ASD genes to identify vulnerable progenitor and neuronal populations and to connect transcriptional regulators to altered fate decisions [13].

Gordon and colleagues compared cortical organoids derived from 70 quality-controlled hiPSC lines from 55 individuals, encompassing genetically defined ASD, idiopathic ASD, and unaffected controls [18]. Across four developmental time points, early expression changes differed by mutation, while the genetically defined ASD groups showed increasing transcriptional convergence with maturation. No significant shared signal was detected across the idiopathic cases in this cohort. This finding limits the study’s evidence for convergence; it does not rule out shared mechanisms in other idiopathic ASD cohorts or under different assays. The CRISPR validation used induced human neural progenitor cells rather than the organoid cohort.

Disease-specific organoids also permit mechanistic analysis of individual genes. SHANK3-deficient single-rosette-derived telencephalic organoids offer a structurally controlled system for studying synaptic and transcriptional consequences of SHANK3 deficiency [16]. Lin and colleagues extended ASD modeling to spatial organization: single-cell and spatial transcriptomics of patient-derived cortical organoids showed local, variably distributed disruptions of progenitor–neuron architecture [27]. Gene-specific models provide depth and isogenic control; pooled screens provide breadth; multi-line cohorts test generalizability; and spatial profiling preserves local tissue context.

6.2. Microcephaly and forebrain growth

Microcephaly provides a particularly direct link between cellular mechanisms and tissue-scale phenotype. Lancaster and colleagues’ foundational cerebral organoid study used patient-derived cells with CDK5RAP2-associated microcephaly and observed premature neuronal differentiation and depletion of progenitor populations, connecting altered progenitor behavior with reduced organoid growth [2]. Qian and colleagues later used region-specific forebrain organoids to model Zika virus exposure, demonstrating impaired neural progenitor development and tissue growth in a human three-dimensional system [20]. These examples show that both genetic and environmental insults can converge on progenitor depletion or disrupted neurogenesis.

PTEN dosage further illustrates how growth phenotypes depend on perturbation direction and genetic background. Mild PTEN overexpression produces smaller organoids [17]. By contrast, the heterozygous PTEN p.Ile135Leu variant increased cortical organoid size at week 4 in both control and ASD genetic backgrounds. Subsequent growth and neurodevelopmental effects depended on background: enlargement persisted at week 8 in the control background but not in the ASD background, while increased progenitor and neuronal production was observed in the ASD background but not in the control background [28]. This comparison supports dosage- and context-sensitive interpretation rather than a universal growth effect for all PTEN variants.

High-throughput screening can broaden the set of mechanisms linked to brain size. Esk and colleagues used three-dimensional human neural tissue to screen candidate microcephaly genes and identify pathways regulating organoid growth [19]. Wang and colleagues added a larger patient-derived phenotypic atlas, including microcephaly and other NDD categories; histology and single-cell profiles from more than 6,000 organoids represented 35 patients and 10 healthy controls, so donor-level replication remains distinct from organoid-level sampling [26]. Comparative organoid work can also illuminate species-specific expansion. Otani and colleagues compared human and chimpanzee cortical progenitor behavior and found differences in proliferation and cell-cycle dynamics that contribute to divergent cortical expansion [21]. These studies link genetic, environmental, and developmental mechanisms to growth phenotypes without reducing microcephaly to a single model state.

7. Limitations, Reproducibility, Ethics, and Future Directions

7.1. Experimental design and reproducibility

An isogenic comparison aims to keep the remaining genetic background constant while testing a defined change at the edited locus, but it is not sufficient by itself to establish reproducibility. Neural organoid studies should distinguish independent donors or cell lines, independent differentiation experiments, replicate culture batches, and multiple organoids sampled within each batch. Organoids generated within the same differentiation are nested biological subsamples; treating every organoid as a fully independent replicate can inflate the apparent sample size and create pseudoreplication. The experimental unit should therefore match the level at which the independent biological manipulation was performed, and statistical models should account for donor and batch structure when appropriate [25].

Replication across independent differentiations is especially important because stochastic patterning, starting-cell quality, matrix lots, media preparation, and operator differences can alter developmental outcomes. Reporting should include the number of donors or cell lines, the number of independent differentiations, the number of organoids analyzed per condition within each differentiation, exclusion or quality-control criteria, and whether analyses were blinded. When genome editing is used, multiple independent clones or complementary rescue experiments can reduce the risk that a phenotype reflects clonal variation or off-target effects rather than the intended perturbation.

Culture-related stress is another confounder. Bhaduri and colleagues demonstrated that cortical organoids can activate stress programs that distort molecular subtype specification [10]. For this reason, disease-related transcriptional changes should be compared not only with isogenic or unaffected controls but also with markers of hypoxia, glycolytic stress, apoptosis, or other culture artifacts. Primary developing human brain datasets provide a useful reference for cell identity and developmental timing, but they are not perfect “ground truth”: tissue source, gestational age, dissociation, preservation, and assay platform can also introduce differences. Benchmarking is strongest when multiple reference datasets and orthogonal measurements agree.

7.2. Maturation and model boundaries

Most neural organoids reproduce selected fetal or early postnatal features rather than the full architecture and physiology of a mature human brain. Long-term culture can extend molecular and electrophysiological maturation [9], and vascular-like engineering can improve support for larger tissue [22], but challenges remain in vascular perfusion, immune integration, myelination, long-range connectivity, and reproducible late-stage cell states. Claims should therefore be tied to the specific developmental feature measured rather than described globally as “brain-like.”

These limits also constrain translation. Organoids are already valuable for mechanistic experiments, genotype-phenotype studies, pathway discovery, and preclinical perturbation or compound screening. In contrast, routine personalized diagnostics, transplantation-based repair, regenerative medicine, and “clinical-grade” neural organoid platforms remain longer-term possibilities. Progress toward those applications will require stronger validation across cohorts, scalable and standardized manufacturing, quality-control benchmarks, safety testing, and regulatory frameworks.

7.3. Ethics and oversight

Current neural organoids do not provide evidence of organism-level sentience or consciousness, and electrophysiological oscillations should not be equated with conscious experience. Nevertheless, ethical analysis should distinguish present evidence from future concerns. As models become more complex, questions may arise about thresholds of functional complexity, donor consent for sensitive neural models, genome editing, transplantation into animals, and the interpretation of increasingly sophisticated neural activity [23].

Oversight should be proportionate to the intervention. Standard in vitro organoid experiments require the usual institutional controls for human pluripotent stem-cell research, biosafety, and responsible data use. Projects involving substantial genome editing require careful assessment of the scientific rationale and genetic manipulation. Transplantation of human neural tissue into animals raises additional questions concerning animal welfare, chimerism, and functional integration. The ISSCR Guidelines for Stem Cell Research and Clinical Translation provide an international framework for scientific rigor, transparency, review, and responsible clinical translation; the 2025 update was targeted to stem-cell-based embryo models, while the other guideline sections remained as published in 2021 [24].

The purpose of this oversight is not to assume that current organoids possess moral status equivalent to a human brain. Rather, it is to ensure that governance develops alongside technical capability. Separating demonstrated properties from hypothetical future capacities avoids both ethical complacency and sensational overstatement.

8. Conclusion

Human brain organoids have become powerful experimental models for studying selected processes in neurodevelopment, but their scientific value depends on matching model architecture to the question being asked. Unguided cerebral organoids provide broad cellular diversity, guided region-specific systems improve control, single-rosette models reduce specific sources of architectural variability, and assembloids recreate interactions between distinct developmental compartments. These approaches should be compared according to reproducibility, complexity, maturation, and intended use rather than ranked as universally better or worse.

The integration of organoids with single-cell transcriptomics, chromatin accessibility, spatial measurements, and CRISPR-based perturbation has shifted the field from descriptive modeling toward causal analysis. Direct organoid multi-omics reconstructs regulatory programs [11], integrated and patient-derived atlases benchmark fidelity and disease phenotypes [12,26], spatial profiling resolves focal architecture [27], and pooled screens test disease-associated genes in developmental contexts [7,13,19]. In ASD, Gordon and colleagues found increasing transcriptional convergence among genetically defined forms, but no significant shared signal across the idiopathic cases in their cohort [18]. In microcephaly, organoids connect genetic dosage, infection, progenitor dynamics, and tissue growth to measurable developmental outcomes [2,17,19,20,28].

At the same time, the field’s major limitations are not solved by technological complexity alone. Reliable interpretation requires independent donors or cell lines, independent differentiations, appropriate organoid-level nesting, careful control of culture stress, and benchmarking against primary human tissue. Ethical discussion likewise requires a clear distinction between what current organoids demonstrably do and what future, more complex systems might make possible. Brain organoids are therefore best understood today as research platforms for mechanistic discovery and perturbation-based testing. Personalized diagnostics, regenerative use, and direct clinical intervention remain prospective goals that will depend on further advances in maturation, validation, scalability, safety, and oversight.

---

Author Contributions

Jihoo Hyun: Conceptualization; Investigation; Writing – original draft; Writing – review and editing.

Conflicts of Interest

The author declares no competing interests.Data Availability Statement

No new data, code, or materials were generated for this narrative review. All information discussed is available in the cited literature.

Preprint disclosure

An earlier version of this manuscript was posted as a preprint in the Research Archive of Rising Scholars (https://doi.org/10.58445/rars.3382). This submission is a revised version of that preprint.

Funding

This work received no external funding.

Ethics statement

Not applicable. This review did not involve human participants, animals, or the collection of primary data.

---

References

1. Pașca, S. P., Arlotta, P., Bateup, H. S., Camp, J. G., Cappello, S., Gage, F. H., et al. (2022). A nomenclature consensus for nervous system organoids and assembloids. Nature, 609, 907-910. https://doi.org/10.1038/s41586-022-05219-6

2. Lancaster, M. A., Renner, M., Martin, C. A., Wenzel, D., Bicknell, L. S., Hurles, M. E., et al. (2013). Cerebral organoids model human brain development and microcephaly. Nature, 501, 373-379. https://doi.org/10.1038/nature12517

3. Chiaradia, I., Imaz-Rosshandler, I., Nilges, B. S., Boulanger, J., Pellegrini, L., Das, R., et al. (2023). Tissue morphology influences the temporal program of human brain organoid development. Cell Stem Cell, 30(10), 1351-1367.e10. https://doi.org/10.1016/j.stem.2023.09.003

4. Tidball, A. M., Niu, W., Ma, Q., Takla, T. N., Walker, J. C., Margolis, J. L., et al. (2023). Deriving early single-rosette brain organoids from human pluripotent stem cells. Stem Cell Reports, 18(12), 2498-2514. https://doi.org/10.1016/j.stemcr.2023.10.020

5. Yoon, S. J., Elahi, L. S., Pașca, A. M., Marton, R. M., Gordon, A., Revah, O., et al. (2019). Reliability of human cortical organoid generation. Nature Methods, 16, 75-78. https://doi.org/10.1038/s41592-018-0255-0

6. Xiang, Y., Tanaka, Y., Patterson, B., Kang, Y. J., Govindaiah, G., Roselaar, N., et al. (2017). Fusion of regionally specified hPSC-derived organoids models human brain development and interneuron migration. Cell Stem Cell, 21, 383-398.e7. https://doi.org/10.1016/j.stem.2017.07.007

7. Meng, X., Yao, D., Imaizumi, K., et al. (2023). Assembloid CRISPR screens reveal impact of disease genes in human neurodevelopment. Nature, 622, 359-366. https://doi.org/10.1038/s41586-023-06564-w

8. Trujillo, C. A., Gao, R., Negraes, P. D., Gu, J., Buchanan, J., Preissl, S., et al. (2019). Complex oscillatory waves emerging from cortical organoids model early human brain network development. Cell Stem Cell, 25, 558-569.e7. https://doi.org/10.1016/j.stem.2019.08.002

9. Gordon, A., Yoon, S. J., Tran, S. S., Makinson, C. D., Park, J. Y., Andersen, J., et al. (2021). Long-term maturation of human cortical organoids matches key early postnatal transitions. Nature Neuroscience, 24, 331-342. https://doi.org/10.1038/s41593-021-00802-y

10. Bhaduri, A., Andrews, M. G., Mancia Leon, W., Jung, D., Shin, D., Allen, D., et al. (2020). Cell stress in cortical organoids impairs molecular subtype specification. Nature, 578, 142-148. https://doi.org/10.1038/s41586-020-1962-0

11. Fleck, J. S., Jansen, S. M. J., Wollny, D., et al. (2023). Inferring and perturbing cell fate regulomes in human brain organoids. Nature, 621, 365-372. https://doi.org/10.1038/s41586-022-05279-8

12. He, Z., Dony, L., Fleck, J. S., et al. (2024). An integrated transcriptomic cell atlas of human neural organoids. Nature, 635, 690-698. https://doi.org/10.1038/s41586-024-08172-8

13. Li, C., Fleck, J. S., Martins-Costa, C., et al. (2023). Single-cell brain organoid screening identifies developmental defects in autism. Nature, 621, 373-380. https://doi.org/10.1038/s41586-023-06473-y

14. Antón-Bolaños, N., Faravelli, I., Faits, T., et al. (2024). Brain Chimeroids reveal individual susceptibility to neurotoxic triggers. Nature, 631, 142-149. https://doi.org/10.1038/s41586-024-07578-8

15. Paulsen, B., Velasco, S., Kedaigle, A. J., Pigoni, M., Quadrato, G., Deo, A. J., et al. (2022). Autism genes converge on asynchronous development of shared neuron classes. Nature, 602, 268-273. https://doi.org/10.1038/s41586-021-04358-6

16. Wang, Y., Chiola, S., Yang, G., et al. (2022). Modeling human telencephalic development and autism-associated SHANK3 deficiency using organoids generated from single neural rosettes. Nature Communications, 13, 5688. https://doi.org/10.1038/s41467-022-33364-z

17. Dhaliwal, N., Choi, W. W. Y., Muffat, J., & Li, Y. (2021). Modeling PTEN overexpression-induced microcephaly in human brain organoids. Molecular Brain, 14, 131. https://doi.org/10.1186/s13041-021-00841-3

18. Gordon, A., Yoon, S. J., Bicks, L. K., Martín, J. M., Pintacuda, G., Arteaga, S., et al. (2026). Developmental convergence and divergence in human stem cell models of autism. Nature, 651, 707-719. https://doi.org/10.1038/s41586-025-10047-5

19. Esk, C., Lindenhofer, D., Haendeler, S., et al. (2020). A human tissue screen identifies a regulator of ER secretion as a brain-size determinant. Science, 370, 935-941. https://doi.org/10.1126/science.abb5390

20. Qian, X., Nguyen, H. N., Song, M. M., Hadiono, C., Ogden, S. C., Hammack, C., et al. (2016). Brain-region-specific organoids using mini-bioreactors for modeling ZIKV exposure. Cell, 165, 1238-1254. https://doi.org/10.1016/j.cell.2016.04.032

21. Otani, T., Marchetto, M. C., Gage, F. H., Simons, B. D., & Livesey, F. J. (2016). 2D and 3D stem cell models of primate cortical development identify species-specific differences in progenitor behavior contributing to brain size. Cell Stem Cell, 18, 467-480. https://doi.org/10.1016/j.stem.2016.03.003

22. Cakir, B., Xiang, Y., Tanaka, Y., Kural, M. H., Parent, M., Kang, Y. J., et al. (2019). Engineering of human brain organoids with a functional vascular-like system. Nature Methods, 16, 1169-1175. https://doi.org/10.1038/s41592-019-0586-5

23. Farahany, N. A., Greely, H. T., Hyman, S., Koch, C., Grady, C., Pașca, S. P., et al. (2018). The ethics of experimenting with human brain tissue. Nature, 556, 429-432. https://doi.org/10.1038/d41586-018-04813-x

24. International Society for Stem Cell Research. (2025). Guidelines for Stem Cell Research and Clinical Translation, updated 2025. ISSCR. https://www.isscr.org/guidelines

25. Sandoval, S. O., Cappuccio, G., Kruth, K., Osenberg, S., Khalil, S. M., Méndez-Albelo, N. M., et al. (2024). Rigor and reproducibility in human brain organoid research: Where we are and where we need to go. Stem Cell Reports, 19(6), 796-816. https://doi.org/10.1016/j.stemcr.2024.04.008

26. Wang, L., Nakamura, Y., Li, J., et al. (2025). A phenotypic brain organoid atlas and biobank for neurodevelopmental disorders. Cell Stem Cell, 32(12), 1923-1940.e7. https://doi.org/10.1016/j.stem.2025.10.006

27. Lin, L., Saw, T. Y., Chou, N., et al. (2026). Spatial architecture of autism pathogenesis reveals mosaic structural disarray during early development. Nature Communications, 17, 7522. https://doi.org/10.1038/s41467-026-74320-5

28. Fu, S., Bury, L. A. D., Eum, J., & Wynshaw-Boris, A. (2023). Autism-specific PTEN p.Ile135Leu variant and an autism genetic background combine to dysregulate cortical neurogenesis. American Journal of Human Genetics, 110(5), 826-845. https://doi.org/10.1016/j.ajhg.2023.03.015