Skip to content

NAMs search vocabulary: terms, lineages and application domains

Written 2026-10-20 for issue #14428, in support of the NAMs landscape analysis. "New Approach Methodologies" (NAMs) is used in more than one sense, so before DisMech is searched for NAM content it helps to pin down what the term covers. This report sets out the search terms that describe NAMs, where the concept came from, and the uses it is put to. Its purpose is to decide what to look for in DisMech.

This report does not map anything to DisMech yet. It contains no counts of DisMech entries and makes no claim about what DisMech does or does not hold. The last section lists where a mapping would start.

Sources. OpenScientist job e7bed1f0-eb28-4981-aace-6559505b2ad0 (run 2026-10-15) supplied the initial list of search terms, and a follow-up chat on that job supplied the lineages and application domains. The list was then expanded with five terms (in vitro, complex in vitro, precision medicine, patient-specific and real-world evidence), and every count was re-run against PubMed on 2026-10-19 with scripts/nams_term_counts.py. The lineages, their dates and the application domains are OpenScientist's statements and have not been checked against primary sources, with one exception noted in section 2.

Summary

  • No single term retrieves the NAMs literature. "New approach methodologies" is not a MeSH heading (a MeSH search for it returns no records), and OpenScientist found the phrase itself is recent. Most NAM papers name a method (organoid, organ-on-chip, QSAR, PBPK) without using the umbrella term, so a search has to combine many method terms.
  • The NAMs concept has three lineages: regulatory toxicology, FDA drug and biologics development, and computational-clinical simulation. A DisMech search that follows only one of them will miss NAM content from the others.
  • NAMs are used in six application domains. Two of them, precision medicine and real-world evidence, were singled out as critical; three of the added search terms cover them.
  • Term counts span four orders of magnitude. In vitro alone matches 1,669,467 PubMed records and in silico 107,148, while specialist terms such as tissue chip match a few hundred. The broad terms need a filter before they say anything about NAMs.

1. Search terms and PubMed counts

Each count is the number of PubMed records whose title or abstract matches the search, on 2026-10-19. Terms overlap, so the counts do not add up to a total. IATA (246) and FDA Modernization Act (136) were counted but are left out of the list.

What the * means. The labels with an asterisk are OpenScientist's. For organoid* and spheroid* the asterisk is PubMed's truncation wildcard: the search matches every word that starts with those letters. For the seven labels marked † the search used no wildcard; it listed exact phrases instead, usually the singular and the plural. For example, digital twin* was searched as "digital twin"[tiab] OR "digital twins"[tiab].

Term Search run (title/abstract) PubMed records
in vitro "in vitro"[tiab] 1,669,467
in silico "in silico"[tiab] 107,148
patient-specific "patient-specific"[tiab] OR "patient specific"[tiab] 44,242
precision medicine "precision medicine"[tiab] 42,308
iPSC / hiPSC "induced pluripotent stem cell"[tiab] OR "induced pluripotent stem cells"[tiab] OR iPSC[tiab] OR hiPSC[tiab] 34,792
organoid* organoid*[tiab] 27,786
spheroid* spheroid*[tiab] 26,627
high-throughput screening "high-throughput screening"[tiab] OR "high throughput screening"[tiab] 24,713
QSAR QSAR[tiab] OR "quantitative structure-activity relationship"[tiab] 16,020
real-world evidence "real-world evidence"[tiab] OR "real world evidence"[tiab] 12,901
3D bioprint* † "3D bioprinting"[tiab] OR bioprinted[tiab] OR bioprinting[tiab] 8,163
PBPK / PBK PBPK[tiab] OR PBK[tiab] OR "physiologically based pharmacokinetic"[tiab] OR "physiologically based kinetic"[tiab] 7,407
adverse outcome pathway* † "adverse outcome pathway"[tiab] OR "adverse outcome pathways"[tiab] OR AOP[tiab] 4,714
digital twin* † "digital twin"[tiab] OR "digital twins"[tiab] 4,161
organ-on-chip* † "organ-on-a-chip"[tiab] OR "organ-on-chip"[tiab] OR "organ on a chip"[tiab] OR "organs-on-chips"[tiab] 3,748
virtual patient / cohort "virtual patient"[tiab] OR "virtual patients"[tiab] OR "virtual cohort"[tiab] 2,692
microphysiological system* † "microphysiological system"[tiab] OR "microphysiological systems"[tiab] 1,607
complex in vitro "complex in vitro"[tiab] 1,266
read-across "read-across"[tiab] OR "read across"[tiab] 1,050
ToxCast / Tox21 ToxCast[tiab] OR Tox21[tiab] 855
IVIVE / QIVIVE IVIVE[tiab] OR QIVIVE[tiab] OR "in vitro to in vivo extrapolation"[tiab] 819
alternatives to animal testing "alternatives to animal testing"[tiab] OR "alternative to animal testing"[tiab] 517
tissue chip* † "tissue chip"[tiab] OR "tissue chips"[tiab] 380
HTTr "high-throughput transcriptomics"[tiab] OR "high throughput transcriptomics"[tiab] OR HTTr[tiab] 273
in silico clinical trial* † "in silico clinical trial"[tiab] OR "in silico clinical trials"[tiab] OR "in silico trial"[tiab] 262

Two caveats on individual searches:

  • The adverse outcome pathway* search also matches the abbreviation AOP, which has other meanings, so that count may include unrelated records.
  • In vitro, patient-specific and precision medicine count every PubMed record that uses them, not only NAM papers.

The full table, with the counts OpenScientist reported on 2026-10-15 beside the current ones, is data/nams-term-counts-2026-10-19.tsv. The 20 terms from the original list changed by at most 0.6% between the two dates.

2. Lineages and application domains

The lineages are the historical sources of NAM methods. The application domains are the places NAM outputs are used.

flowchart LR
    L1["Regulatory toxicology<br/>1980s–2000s"] --> N
    L2["FDA drug and biologics development<br/>FDA Modernization Act 2.0, Dec 2022"] --> N
    L3["Computational-clinical simulation<br/>2010s onward"] --> N
    N(("NAMs today"))
    N --> D1["Precision medicine and<br/>pharmacogenomics (critical)"]
    N --> D2["Real-world evidence<br/>(critical)"]
    N --> D3["Environmental risk assessment"]
    N --> D4["Cosmetics and food safety"]
    N --> D5["Biomarker-driven trial design"]
    N --> D6["Regulatory pathway<br/>optimization (IND/NDA)"]
Lineage Origins Purpose Characteristic methods
Regulatory toxicology (1980s–2000s) EPA, NICEATM/ICCVAM, Tox21/ToxCast, NIEHS/NTP Chemical safety assessment with fewer animal tests Adverse outcome pathways, read-across, high-throughput screening
FDA drug and biologics development (organ chips from the 2010s; law signed December 2022) FDA Modernization Act 2.0; FDCA §505(z) "nonclinical tests" Human-relevant prediction of drug safety and efficacy Organoids, organ chips, iPSC-derived systems, PBPK, QSAR
Computational-clinical simulation (2010s, accelerating in the 2020s) Clinical pharmacometrics, quantitative systems pharmacology, model-informed drug development Human dose prediction and trial design Digital twins, virtual patients and cohorts, in silico clinical trials

The third lineage came from the follow-up chat. The 10-iteration run described NAMs as having two lineages. The computational-clinical lineage was named afterwards, from terms the run had grouped as one of its search clusters.

FDA Modernization Act 2.0 date. The chat dates the Act to 2023. It was signed into law on 29 December 2022, as stated in the abstract of Han 2023, PMID:36762462. The article was published in 2023, which is the likely source of the 2023 date.

The six application domains come from the chat's proposed taxonomy. It names precision medicine and real-world evidence as the two critical layers: precision medicine for patient stratification and individual prediction, and real-world evidence as data that validates or contextualises NAM predictions.

3. Search terms by application domain

This grouping is a draft, not OpenScientist's. The chat names the six domains but does not assign search terms to them. A term can sit in more than one domain, so the counts within a domain overlap. The Why column gives the reason for each assignment, so it can be corrected term by term.

Term Records Precision medicine Real-world evidence Environmental risk Cosmetics and food Trial design IND/NDA Why
in vitro 1,669,467 ● Cell-based assays are one of the nonclinical test types FDCA §505(z) lists. Too generic to place in a narrower domain.
in silico 107,148 ● Computer modelling is one of the nonclinical test types FDCA §505(z) lists for IND/NDA packages. Too generic to place in a narrower domain.
patient-specific 44,242 ● The follow-up chat lists "patient-specific" as a precision-medicine term the vocabulary missed.
precision medicine 42,308 ● Names the precision-medicine domain directly.
iPSC / hiPSC 34,792 ● Patient-derived iPSC systems are the chat's first example of a precision-medicine use.
organoid* 27,786 ● Patient-derived organoids are used to test individual drug response.
spheroid* 26,627 ● Patient-derived tumour spheroids are used for individual drug-response testing.
high-throughput screening 24,713 ● The Tox21/ToxCast route for ranking chemical hazard.
QSAR 16,020 ● ● Predicts chemical hazard; used for environmental chemicals and for cosmetic ingredients where animal tests are restricted.
real-world evidence 12,901 ● Names the real-world evidence domain directly.
3D bioprint* † 8,163 ● Bioprinted and reconstructed skin models are used for cosmetic irritation and sensitisation testing.
PBPK / PBK 7,407 ● ● Drives dose selection in trial design and is part of model-informed drug development in IND/NDA submissions.
adverse outcome pathway* † 4,714 ● ● The organising framework for chemical risk assessment, including skin sensitisation for cosmetics.
digital twin* † 4,161 ● ● The chat pairs digital twins with personalised prediction, and they are used to design trial arms.
organ-on-chip* † 3,748 ● ● Patient-derived chips for individual response, and qualification as drug development tools.
virtual patient / cohort 2,692 ● Used to simulate trial populations and stratify them.
microphysiological system* † 1,607 ● ● The chat names patient-derived MPS for precision medicine. The IQ MPS Consortium works on regulatory qualification.
complex in vitro 1,266 ● Complex in vitro models (organoids, organ chips, MPS) are the term used in drug-safety and regulatory qualification discussions.
read-across 1,050 ● ● Fills data gaps for untested chemicals, including cosmetic ingredients.
ToxCast / Tox21 855 ● EPA and NCATS chemical-screening programmes.
IVIVE / QIVIVE 819 ● ● Converts in vitro concentrations to human doses, for chemical risk and for drug submissions.
alternatives to animal testing 517 ● The framing used by cosmetics regulation, where animal testing is banned in several markets.
tissue chip* † 380 ● The NCATS Tissue Chip programme term, aimed at drug-development use.
HTTr 273 ● EPA high-throughput transcriptomics for chemical points of departure.
in silico clinical trial* † 262 ● Simulated trials used to design or supplement real ones.

Column key: Precision medicine = precision medicine and pharmacogenomics; Real-world evidence = RWD/RWE integration; Environmental risk = environmental risk assessment and remediation; Cosmetics and food = cosmetics and food safety; Trial design = biomarker-driven clinical-trial design; IND/NDA = regulatory pathway optimization.

Where a mapping to DisMech would start

The mapping itself is the next step and is not done here. NAM content in DisMech is recorded mainly in the experimental_models slot of a disorder entry. Each ExperimentalModel has an experimental_model_type (ORGANOID, ORGAN_ON_CHIP, CELL_LINE, IPSC_DERIVED_MODEL, PRIMARY_CELL_CULTURE, CO_CULTURE or OTHER) and can link to pathograph nodes through modeled_mechanisms. These model types cover the in vitro methods from the FDA lineage. A mapping would also need to look elsewhere for the two other lineages, because their methods are computational or assay-based rather than model systems: adverse outcome pathways, ToxCast/Tox21 assays and read-across from regulatory toxicology, and PBPK, QSP, digital twins and virtual patients from computational-clinical simulation.

Reproducing the counts

uv run python scripts/nams_term_counts.py --out docs/reports/data/nams-term-counts-<date>.tsv

The script re-runs each search exactly as OpenScientist ran it (from the job's provenance/iter4_transcript.json), adds the expanded terms, and prints the current count beside the reported one. PubMed grows daily, so counts drift upward between runs.