AsianScientist (Mar. 05, 2026) –Neurodegenerative disorders are among the most difficult puzzles in biomedicine. Part of the challenge lies in the extraordinary complexity of the human brain and the task of modelling the intricate processes that drive irreversible nerve cell death.
Parkinson’s disease, for example, involves the gradual loss of midbrain dopaminergic neurons (mDA)—cells that produce dopamine, the chemical messenger essential for movement control and learning. Scientists have spent years trying to recreate them in the lab by mimicking the chemical cues that shape their development in the embryo and coaxing stem cells to become midbrain-like, dopamine-producing cells in a dish.
However, differentiation protocols produce widely varying numbers of mDA neurons and it is still uncertain how closely these cells mirror their natural counterparts at the molecular level. If they fall short, disease models could mislead and therapies built on them may not work as intended.
To vet lab-grown midbrain models more rigorously, researchers from Duke-NUS Medical School in the National University of Singapore worked with collaborators from the University of Sydney to build BrainSTEM (Brain Single-Cell Two-tier Mapping).
BrainSTEM provides a detailed, single-cell atlas of the human fetal brain. It maps gene activity in individual cells to create a comprehensive reference of early brain development. The team also included a specialized midbrain subatlas that zooms in on midbrain cell types.
Using this two-tier reference framework, the researchers evaluated current differentiation protocols. Published single-cell RNA sequencing datasets of lab-grown mDA neurons were first mapped onto the whole fetal brain atlas to determine their regional identity. This step helped identify off-target cells that had adopted features from other brain regions. Only cells classified as midbrain-like were then mapped onto the high-resolution midbrain subatlas to detect true mDA neurons.
The analysis showed that while many methods successfully generated authentic midbrain cell types, they also produced off-target populations. Importantly, the study helped identify patterns in protocol performance that could serve as guideposts for improving differentiation strategies.
What sets BrainSTEM apart from traditional validation approaches is mapping cells onto the whole brain first, rather than directly to a fetal midbrain reference. Skipping this step risks misclassifying off-target cells as midbrain types. This layered, high-resolution strategy also enabled the discovery of a rare dopaminergic subpopulation — hDA.STN — that might have otherwise remained obscured.
The team have packaged BrainSTEM into a user-friendly R tool, making it accessible to researchers worldwide. More broadly, the study advocates for multilevel mapping for evaluating complex biological systems, even beyond the midbrain.
“BrainSTEM marks a significant step forward in brain modelling,” said Alfred Sun, an assistant professor from Duke-NUS’ Neuroscience & Behavioural Disorders programme and a senior author of the paper.
“By delivering a rigorous, data-driven approach, it will speed the development of reliable cell therapies for Parkinson’s disease. We’re setting a new standard to ensure the next generation of Parkinson’s models truly reflects human biology.”
—
Source: National University of Singapore ; Image: Vecstock/Freepik
This article can be found at BrainSTEM: A single- cell multiresolution fetal brain atlas reveals transcriptomic fidelity of human midbrain cultures.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
