Most current commentary stops at advocacy: NAMs are human-relevant, therefore use them. That framing is not usable at the bench. A screening cascade is built around specific questions — does the compound engage the target, does it alter synaptic function, does it shift circuit-level activity, does it cross the blood-brain barrier at relevant exposure, and does it move a biomarker that translates to the clinic. Each question favours a different assay, and the evidence quality required differs at each step. FDA’s draft guidance on NAMs in drug development sets out a validation framework and general recommendations; it is a draft and not for implementation, but it makes assay fit and documented validation a study-design problem rather than a procurement one.
What follows compares human iPSC-derived neuronal systems, brain slice electrophysiology, microphysiological systems, and computational models against those five decisions.
There is no ranking of “best” models here — only fit between the biological question and the readout that can answer it. The FDA’s own framing of new approach methodologies spans human-based in vitro systems, in silico modeling, and other innovative platforms, which is a useful reminder that these are complementary tiers, not competing products.
| Question you need answered | Model tier that fits | Typical readouts |
| Does the target exist and behave in human cells? | Human iPSC-derived neurons and glia | Manual patch clamp, calcium imaging, multi-electrode firing |
| Is synaptic transmission or plasticity altered? | Acute or organotypic brain slice | Field potentials, long-term potentiation, evoked and spontaneous currents |
| Does the compound reach the target? | Barrier and microphysiological systems | Permeability, transporter flux, unbound exposure |
| Which of 200 candidates deserve a rig? | In silico and pharmacology modeling | Ranked hypotheses, multi-parameter integration |
Human iPSC-derived cultures give scalable, human-genome-relevant pharmacology, and specialists such as Metrion Biosciences have built stem cell-derived neuronal firing assays specifically for neurotoxicology and seizure liability screening. What those cultures cannot yet reproduce is native circuitry: intact synaptic connectivity, defined afferent pathways, and region-specific architecture. That is where slice electrophysiology remains the reference method for plasticity and network-level questions.
Blood-brain barrier chips and other engineered tissue platforms — the class of systems NCATS funds through its Tissue Chip for Drug Screening program — address exposure and transport, which cell-only assays cannot. Computational triage narrows the list; it does not validate a mechanism. The design skill is deciding which two or three validated assays, at different levels of biology, together de-risk the decision in front of you.
Human induced pluripotent stem cell (iPSC)-derived neurons and glia are the workhorse of most human-relevant new approach methodologies in central nervous system programs. They carry patient genetic background, so they support disease-phenotype characterization, mechanism-of-action work on a defined human target, and compound screening at scale — exactly the class of in vitro human-based systems the FDA names in its description of these methods.
Model choice should follow the clinical question, not the novelty of the preparation. If the hypothesis concerns long-term potentiation, the readout has to be a plasticity measurement; if it concerns a channel’s state-dependent block, it has to be a voltage-clamp protocol. Biomarkers picked after the model is chosen tend to describe the culture rather than the patient.
Human relevance is not predictiveness. The FDA’s NAMs page states that over 90% of drugs found safe in animals fail in humans, so a phenotype worth a go/no-go call needs orthogonal confirmation: manual patch clamp, calcium imaging and high-density multielectrode array recording across both iPSC-derived neurons and native brain slice tissue.
The FDA characterizes new approach methodologies as testing methods and strategies used to assess the safety, efficacy, and quality of drugs, spanning in vitro human-based systems, in silico modeling, and other emerging platforms. Its draft guidance on general considerations for these methods — dated March 2026 and explicitly not for implementation — sets out a validation framework and general recommendations rather than a list of accepted assays. The practical message for study design is that context of use, not platform novelty, determines acceptability.
CDER’s position on streamlined nonclinical studies and acceptable NAMs is similarly conditional: these methods, when combined with pathway knowledge, pharmacology, and other data, may reduce animal use and support the 3Rs. Reduction follows from a coherent evidence package, not from a single human-cell readout.
On the public-research side, NIH’s National Center for Advancing Translational Sciences runs programs including Tissue Chip for Drug Screening, Tox21, the Stem Cell Translation Laboratory, and 3-D Tissue Bioprinting — durable infrastructure that has matured microphysiological systems into credible tools.
For CNS programs, those systems earn their place when the scientific question involves barrier transport, multi-cellular interaction, or sustained tissue-level exposure. They are weaker where the endpoint is millisecond synaptic function or circuit-level plasticity, which is where slice and cell recordings with manual patch clamp remain the sharper instrument.
Regulatory interest raises the documentation bar; it does not substitute for assay-specific validation data.
1. Define the decision before the assay. Write down what the data has to resolve — target validation, hit triage, seizure or neurotoxicity liability, brain exposure, or biomarker alignment with a clinical endpoint. The FDA’s draft guidance on general considerations for new approach methodologies (dated March 2026 and explicitly not for implementation) organizes validation around context of use, so a plan that names its decision point is easier to defend than one that names a platform.
2. Establish human relevance early. Human iPSC-derived neuronal and glial preparations, read with calcium imaging, HD-MEA network activity, and manual patch clamp on the target channel or receptor, give human-genome-relevant pharmacology at screening throughput.
3. Escalate to tissue when the mechanism is synaptic or circuit-level. Brain slice electrophysiology — evoked field potentials, long-term potentiation and other synaptic plasticity measures, spinal dorsal horn recordings for pain programs — captures intact connectivity that dissociated cells cannot report.
4. Add barrier and microphysiological models when transport, efflux, or CNS exposure sits in the risk profile rather than being assumed from plasma data.
5. Triangulate readouts. Pair imaging, electrophysiology, and molecular endpoints so no single assay carries the whole translational burden — the FDA notes these methods are most useful combined with pathway knowledge and pharmacology, and cites the failure of most animal-safe candidates in humans.
Neuroservices-Alliance is a PhD-led contract research organization for central nervous system and pain programs, built around electrophysiology recordings in brain slices and in cells. The readouts — manual patch clamp, calcium imaging, HD-MEA network recordings — earn their place when a program needs mechanistic resolution: is the target engaged at the channel or receptor level, does the compound alter synaptic plasticity, does network firing shift in a disease-relevant circuit.
That resolution comes with a plain tradeoff. Functional recordings will not answer blood-brain barrier permeability, metabolism, or transport questions, and they are not a volume screening funnel on their own. Metrion Biosciences, for instance, is explicit about high-throughput ion channel screening cascades, and PsychoGenics has published on machine-learning-driven behavioral phenotyping. Among new approach methodologies, slice and cell electrophysiology sits downstream of triage and upstream of in vivo commitment.
Assays here are designed per question rather than run off a fixed menu, with data intended to hold up in peer review — the standard the FDA’s draft validation framework points toward.
Select models by the decision in front of you, not by novelty. Ion channel pharmacology in human iPSC-derived neurons resolves target engagement; brain slice electrophysiology interrogates synaptic plasticity and circuit-level function; blood-brain barrier models address exposure; microphysiological systems support longer-term, multicellular questions. None substitutes for another, and a stack is only credible when each assay is validated, paired with orthogonal readouts, and anchored to translational biomarkers — the direction the FDA’s draft guidance on new approach methodologies also points, with its emphasis on a validation framework.
