Research Methodology & Publication Terms

Peptide-relevant disease models include: DIO mice (diet-induced obesity — 60% fat diet for 12-16 weeks, modelling metabolic syndrome/T2D), ob/ob mice (leptin-deficient, spontaneous obesity), db/db mice (leptin receptor-deficient, T2D), STZ mice/rats (streptozotocin-induced beta cell destruction, modelling T1D/beta cell failure), WHHL rabbits (familial hypercholesterolaemia), spontaneously hypertensive rats (SHR, for cardiovascular studies), nude/SCID mice (immunodeficient, for xenograft tumour models), EAE mice (experimental autoimmune encephalomyelitis, modelling MS), and collagen-induced arthritis (CIA, modelling rheumatoid arthritis). For GnRH compound testing: castrated animal models with testosterone pellet implants assess androgen suppression. Critical consideration: peptide receptors may differ between species in expression pattern, affinity, or signalling — limiting direct translatability of animal data to humans.

Case reports follow CARE (CAse REport) guidelines: title (key term 'case report'), abstract, introduction (background context), patient information (demographics, medical history), clinical findings, timeline (chronological events), diagnostic assessment, therapeutic interventions, follow-up and outcomes, and discussion (strengths/limitations, rationale for conclusions, relevant literature). For peptide compounds, published case reports may document: first human use of a research peptide (outside formal clinical trials — ethically complex), unexpected adverse reactions to approved peptide drugs (potentially representing new safety signals), novel off-label applications with documented outcomes, and drug interaction discoveries. The British Medical Journal (BMJ) Case Reports and American Journal of Case Reports are high-volume case report outlets. PubMed indexes case reports with the [case reports] publication type filter.

Major guidelines relevant to peptide therapeutics: ADA/EASD Consensus Report for T2D management (positions GLP-1 RAs as preferred second-line therapy after metformin, with cardiovascular-benefit emphasis), NICE NG28 (UK T2D management), ESC/EAS Cardiovascular Prevention guidelines (incorporate GLP-1 RA cardiovascular evidence), AGA guidelines for obesity management, NICE CG189 (obesity), NCCN guidelines for multiple myeloma (bortezomib, carfilzomib), Endocrine Society guidelines for acromegaly (somatostatin analogues), AUA/ASCO guidelines for prostate cancer (GnRH agonists/antagonists), and ESHRE guidelines for IVF protocols (GnRH antagonists). Guidelines are developed through systematic evidence review, expert panel deliberation, and external consultation. They are updated as new evidence emerges — GLP-1 RA positioning has advanced significantly in diabetes and cardiovascular guidelines following CVOT results.

Cochrane Collaboration operates through review groups specialising in disease areas. Cochrane Reviews follow the most rigorous methodology: comprehensive search strategy (including grey literature), duplicate screening, validated risk of bias assessment (Cochrane RoB 2 tool), GRADE evidence certainty ratings, and mandatory update schedules. Cochrane protocols are published prospectively (registered before the review is conducted, preventing selective outcome reporting). The Cochrane Library publishes over 8,000 reviews. Cochrane Reviews of peptide therapeutics (e.g. GLP-1 RAs for T2D, somatostatin analogues for acromegaly, GnRH agonists for prostate cancer) are regarded as the highest-quality evidence syntheses and are directly referenced by guideline committees (NICE, ADA, EASD) in formulating treatment recommendations.

Financial COI types: direct employment by pharmaceutical company, consulting fees, speaker honoraria, research grants, equity/stock ownership, patent royalties, and expert testimony fees. Non-financial COI: academic advancement pressure, intellectual commitments to a hypothesis, and personal relationships. ICMJE (International Committee of Medical Journal Editors) requires standardised COI disclosure forms with all manuscript submissions. Studies sponsored by pharmaceutical companies show larger effect sizes on average than independently funded studies (known as the funding effect). Cochrane Reviews assess funding source as a potential risk of bias factor. Clinical practice guideline committees increasingly require majority independent (non-conflicted) membership and transparent management of members with COI. For peptide drug publications, disclosing industry relationships is essential for readers to assess potential bias.

Consensus methods include: Delphi technique (iterative anonymous questionnaire rounds to converge on agreement), nominal group technique (structured face-to-face discussion with anonymous voting), and consensus development conference (structured presentation of evidence followed by panel deliberation). Consensus statements are used when: evidence is insufficient for formal guidelines, the topic is emerging or controversial, expert interpretation is needed to bridge evidence gaps, or rapid guidance is needed before formal guidelines can be developed. In the peptide therapeutics space, consensus statements have addressed: optimal management of GLP-1 RA gastrointestinal side effects, use of somatostatin analogues in various NET subtypes, and management of GnRH agonist-related bone loss.

DOIs are managed by the International DOI Foundation (IDF) and administered through registration agencies, primarily Crossref (for academic content) and DataCite (for research data). The DOI resolution system (https://doi.org/10.XXXX/XXXXX) redirects to the current location of the content, providing persistent access even when publisher URLs change. DOI metadata (title, authors, publication date, journal) is stored in the Crossref database and accessible via the Crossref API. PeptideTrace's compound reference sections use DOIs to link directly to source publications — this provides permanent, reliable citations that won't break over time. DOIs are increasingly used for: clinical trial registrations, datasets, preprints, and research software, extending their utility beyond traditional journal articles.

The traditional evidence pyramid (bottom to top): expert opinion/editorials, case reports/case series, cross-sectional studies, case-control studies, cohort studies, randomised controlled trials, systematic reviews/meta-analyses. Modern frameworks add nuance: GRADE (Grading of Recommendations Assessment, Development and Evaluation) rates certainty of evidence as high (further research unlikely to change confidence), moderate (further research likely to change confidence), low (further research very likely to change confidence), or very low (any estimate is very uncertain). GRADE considers: risk of bias, inconsistency across studies, indirectness (applicability), imprecision (wide confidence intervals), and publication bias. For peptide compounds, the evidence hierarchy helps contextualise claims: a research compound with only in vitro data (below case reports on the hierarchy) has fundamentally different evidence than an approved drug with multiple Phase III RCTs and systematic reviews.

Ex vivo preparations maintain tissue architecture, cell-cell interactions, and local microenvironment while allowing direct manipulation and observation. Common ex vivo models include: isolated organ preparations (Langendorff perfused heart for cardiac effects, isolated perfused kidney for renal effects), tissue slices (brain slices for electrophysiology, precision-cut liver slices for metabolism), skin explants (Franz diffusion cells for transdermal peptide delivery studies), and isolated blood vessel rings (isometric tension recording for vasoactive effects). Ex vivo studies of peptide effects on isolated human tissue (obtained from surgical specimens with consent) provide the most directly translatable data. The limitation is tissue viability — most ex vivo preparations remain functional for hours to days, restricting study duration.

Impact Factor (IF) = (citations in year Y to articles published in years Y-1 and Y-2) / (number of citable articles published in years Y-1 and Y-2). Example: if a journal published 200 articles in 2023-2024 and those articles received 2,000 citations in 2025, the 2025 IF = 10.0. Top medical journals: New England Journal of Medicine (~176), The Lancet (~168), JAMA (~120). Relevant specialty journals: Diabetes Care (~16), Obesity Reviews (~8), The Lancet Diabetes & Endocrinology (~44). Impact Factor limitations: it is a journal-level metric (not article-level), skewed by a few highly cited articles, manipulable through editorial practices (self-citation, review articles), and varies across disciplines. Article-level metrics (Altmetric score, citation count, h-index for authors) provide complementary assessments. DORA (Declaration on Research Assessment) recommends against using IF as a primary measure of research quality.

Computational methods in peptide drug development include: molecular docking (predicting peptide-receptor binding modes and binding energy), molecular dynamics (MD) simulations (simulating peptide conformational behaviour in solution over nanosecond-microsecond timescales), structure-activity relationship (SAR) modelling (predicting how sequence modifications affect activity), pharmacophore modelling (identifying key structural features required for receptor binding), ADME prediction (computational models for absorption, distribution, metabolism, excretion), and machine learning/AI (training models on peptide activity datasets to predict properties of novel sequences). AlphaFold (protein structure prediction) has accelerated understanding of peptide-receptor interactions. In silico methods reduce experimental screening burden by prioritising the most promising candidates for synthesis and testing. However, computational predictions always require experimental validation.

Common in vitro assays for peptide compounds include: receptor binding assays (radioligand competition — measuring IC50/Ki for receptor affinity), functional assays (cAMP accumulation, calcium flux, reporter gene — measuring EC50/potency), cell viability assays (MTT/MTS, live/dead staining — assessing cytotoxicity), migration/invasion assays (wound healing scratch assay, transwell migration — relevant to tissue repair peptides), and stability assays (plasma stability, microsomal stability, simulated gastric/intestinal fluid stability). In vitro data provide mechanistic understanding and are essential for early drug discovery, but their predictive value for in vivo behaviour is limited by: absence of systemic PK (absorption, distribution, metabolism, excretion), absence of intact physiological feedback systems, and potential for cell culture artefacts. The translational gap between in vitro and in vivo results is a major challenge in peptide drug development.

In vivo peptide research uses various animal species: mice (most common — short gestation, genetic manipulation tools, disease models including ob/ob and db/db diabetic mice, DIO diet-induced obesity), rats (larger size for PK sampling, well-characterised models), rabbits (ophthalmology, immunogenicity), dogs (cardiovascular safety, oral absorption), pigs (skin wound healing — similar to human skin), and non-human primates (closest to human physiology — used when other species lack relevant receptors). Regulatory requirements: IND-enabling toxicology studies typically require one rodent and one non-rodent species. For peptide drugs targeting human-specific receptors, transgenic animals expressing the human receptor may be needed. In vivo PK studies establish bioavailability, half-life, and clearance; PD studies establish dose-response relationships and duration of effect; safety studies identify target organs of toxicity and NOAEL.

CRISPR/Cas9 and earlier technologies (homologous recombination in ES cells) enable targeted gene deletion. Knockout models relevant to peptide pharmacology include: GLP-1 receptor knockout (GLP-1R-/-) mice (used to confirm GLP-1R-dependent effects of GLP-1 RAs — effects abolished in KO mice confirm on-target activity), MC4R knockout mice (develop obesity, modelling MC4R-deficiency obesity targeted by setmelanotide), GHSR knockout mice (resistant to ghrelin's GH-stimulating and orexigenic effects), and various neuropeptide receptor knockouts (elucidating the physiological roles of individual neuropeptide systems). Conditional knockouts (gene deletion in specific tissues or at specific developmental stages using Cre-lox technology) enable more nuanced analysis. Phenotyping of knockout animals reveals the physiological role of the deleted gene/receptor system.

Common classification systems: Oxford Centre for Evidence-Based Medicine (OCEBM) levels 1-5, where Level 1 = systematic review of RCTs or single large RCT with narrow CI, Level 2 = smaller RCTs or large observational studies, Level 3 = case-control/cohort studies, Level 4 = case series/case reports, Level 5 = expert opinion. GRADE system (used by WHO, NICE, and Cochrane) takes a different approach, starting with study design (RCTs start at high, observational studies start at low) and adjusting up or down based on quality factors. For clinical guideline recommendations: Level 1 evidence + strong clinical benefit → strong recommendation (Grade A); lower evidence levels → weaker recommendations (Grade B, C, D). Most research peptides have only Level 4-5 evidence (case reports, expert opinion, extrapolation from preclinical data), while approved peptide drugs are supported by Level 1-2 evidence from clinical trial programmes.

MEDLINE (Medical Literature Analysis and Retrieval System Online) is the NLM's premier bibliographic database, containing over 28 million references to journal articles from approximately 5,200 journals selected for biomedical relevance. MEDLINE's distinguishing feature is MeSH (Medical Subject Headings) indexing — trained indexers assign standardised subject terms to each article, enabling precise topic-based searching. MeSH terms relevant to peptide drug searching include: 'Peptides/therapeutic use', 'GLP-1 Receptor Agonists', 'Somatostatin/analogs & derivatives', specific compound names as MeSH headings or supplementary concepts. MEDLINE's journal selection criteria ensure a baseline quality standard. Journals must pass a scientific quality review by the Literature Selection Technical Review Committee. MEDLINE is accessed primarily through PubMed, which adds non-MEDLINE content and enhanced search features.

Meta-analytical methods: fixed-effects models (assumes one true effect size underlying all studies — appropriate when heterogeneity is low), random-effects models (assumes a distribution of true effects — appropriate when studies have methodological or population differences, producing wider confidence intervals), and Bayesian meta-analysis (incorporates prior knowledge). Heterogeneity assessment: I² statistic (0% = no heterogeneity, 25% = low, 50% = moderate, 75% = high), Cochran's Q test, and prediction intervals (range of effects expected in future studies). Publication bias detection: funnel plot asymmetry, Egger's test, and trim-and-fill analysis. For GLP-1 RA meta-analyses, heterogeneity often arises from: different comparators, varying baseline HbA1c/BMI, different dose levels, and varying treatment durations. Subgroup analyses and meta-regression can explore sources of heterogeneity.

Narrative reviews (also called traditional reviews or scoping reviews) provide expert-synthesised overviews of a topic. They differ from systematic reviews in: search strategy (may not be comprehensive or reproducible), study selection (author-selected rather than criteria-based), quality assessment (may not formally assess bias), and synthesis (qualitative rather than quantitative). Narrative reviews are useful for: providing broad overviews of emerging fields (e.g. 'research peptides in tissue repair'), contextualising clinical data within biological frameworks, educational purposes, and hypothesis generation. However, they are susceptible to selection bias (authors may preferentially cite studies supporting their perspective) and should not be relied upon as definitive evidence for clinical decision-making. Major medical journals publish both narrative and systematic reviews, but regulatory bodies preferentially weight systematic reviews.

NMA (also called mixed treatment comparison, MTC) creates a network of evidence connecting treatments through direct comparisons (A vs B head-to-head) and indirect comparisons (A vs C and B vs C → indirect comparison of A vs B). NMA requires: the transitivity assumption (populations across studies are similar enough for indirect comparisons to be valid) and consistency (direct and indirect estimates agree). Statistical approaches: frequentist (graph-theoretical methods) or Bayesian (Markov chain Monte Carlo — MCMC — modelling). NMA enables ranking of treatments using SUCRA (Surface Under Cumulative Ranking Curve) or P-score values. For GLP-1 RAs, NMAs comparing semaglutide vs tirzepatide vs liraglutide vs dulaglutide vs exenatide provide comprehensive efficacy and safety rankings informing clinical guidelines and health technology assessments (NICE, ICER).

The peer review process: author submits manuscript → editor screens for suitability → 2-3 independent reviewers evaluate methodology, data analysis, conclusions, and novelty → reviewers recommend accept, minor revisions, major revisions, or reject → editor makes decision → revisions incorporated → re-review if needed → acceptance → publication. Single-blind (reviewers know author identity), double-blind (both identities concealed), and open peer review (identities disclosed) formats exist. Peer review limitations: it cannot detect all errors or fraud, reviewers may have biases, and the process takes 2-12 months. PeptideTrace's PubMed-based publication tracking counts peer-reviewed articles indexed with each compound name. Publication count is a proxy for research interest intensity but does not indicate evidence quality — a compound with 500 publications including multiple negative trials has a very different evidence profile than one with 500 positive preclinical reports.

Post hoc analyses are conducted after unblinding and examining results — the knowledge of which treatments succeeded biases which questions are asked. This creates multiplicity concerns: testing many hypotheses increases the probability of finding false positives by chance. Post hoc results should be considered hypothesis-generating rather than confirmatory. They require confirmation in prospective, pre-specified analyses (either in new trials or in pre-registered analyses of existing data). Notable post hoc analyses in peptide therapeutics: subgroup analyses of CVOT data by baseline characteristics (which patient subgroups benefit most from GLP-1 RA cardiovascular protection?), and exploratory analyses of body composition changes with weight management therapies. Post hoc findings can guide Phase IV trial design and inform future research directions.

Sample size formula for a superiority trial comparing two means: n per group = 2 × (Zα/2 + Zβ)² × σ² / δ², where Zα/2 = 1.96 for two-sided α = 0.05, Zβ = 0.84 for 80% power (or 1.28 for 90%), σ = standard deviation of the endpoint, and δ = minimum clinically important difference to detect. Example: to detect a 1.5% HbA1c difference (δ) with SD = 1.2% (σ) at 80% power (two-sided α = 0.05): n = 2 × (1.96+0.84)² × 1.44 / 2.25 = 2 × 7.84 × 0.64 = 10 per group. After adjusting for expected dropout (~20%): approximately 13 per group. For cardiovascular outcomes trials with event rates of 3-5%/year, much larger samples are needed (5,000-17,000 patients) because the event rate is the limiting factor. Power calculations must be documented in the protocol and statistical analysis plan.

Major preprint servers: bioRxiv (biology, operated by Cold Spring Harbor Laboratory), medRxiv (clinical/health sciences, operated by BMJ/Yale/CSHL), arXiv (physics/mathematics/computer science), and ChemRxiv (chemistry). Preprints typically receive a DOI and are permanently archived. They may undergo basic screening (not formal peer review) — bioRxiv screens for: scientific merit (not blank/spam), appropriate subject area, and no patient data/clinical trial results that bypass regulatory oversight. Preprint advantages: rapid dissemination (days vs months for peer review), priority establishment, and community feedback. Limitations: no independent quality validation, potential for errors to be widely cited before correction, and risk of media sensationalisation of unvalidated findings. Some journals screen preprints for articles to invite (accelerating the publication of significant work).

Publication bias distorts the evidence base because: positive trials are 2-3× more likely to be published than negative trials, positive trials are published faster (median 4-5 years for positive vs 6-8 years for negative), and positive trials are more likely to appear in high-impact journals. Detection methods: funnel plot visual inspection (asymmetry suggests bias — missing small negative studies in the lower-left corner), Egger's regression test (formal statistical test for asymmetry), and Begg's rank correlation test. Mitigation strategies: trial registration (ensures awareness of all conducted trials, including unpublished ones), results reporting mandates (FDAAA 801 requires posting results to ClinicalTrials.gov), and journals accepting registered reports (peer review of methods before results are known). For peptide compounds, publication bias is particularly concerning for research compounds where a small number of publications from a limited number of groups may overrepresent positive findings.

PubMed (pubmed.ncbi.nlm.nih.gov) indexes approximately 36 million citations from approximately 5,200 biomedical journals. It accesses: MEDLINE (NLM's curated database with MeSH indexing), PubMed Central (PMC — full-text open access archive), and additional content (ahead-of-print, in-process citations, non-MEDLINE publisher submissions). PeptideTrace uses PubMed's E-utilities API (eutils.ncbi.nlm.nih.gov) — specifically ESearch (search and retrieve record counts) and EFetch (retrieve citation details). The PubMed sync pipeline queries all 185 compounds every 23 hours using quoted exact matching ('CompoundName'[tiab]) to prevent Automatic Term Mapping inflation. Publication counts are stored in compound_research_activity with breakdowns by study type. Important caveat: publication count reflects research interest volume, not evidence quality — a compound may have many publications but limited high-quality clinical evidence.

RUO products are exempt from the regulatory requirements (GMP manufacturing, clinical trials, marketing authorisation) that apply to therapeutic drugs. However, this exemption is conditional: the product must genuinely be intended and sold for research purposes only. Regulatory enforcement targets: products marketed as RUO but sold in quantities, packaging, or through channels clearly intended for human self-administration; products with marketing materials containing therapeutic claims (disease treatment, symptom relief); and websites providing dosing, reconstitution, and injection instructions alongside RUO labelling. The RUO classification exists in regulatory frameworks for diagnostics (21 CFR 809.10(c) for IVD products) but is applied analogously to research chemicals. For peptide compounds, the RUO label is the primary legal mechanism enabling sale of compounds that have not undergone clinical trials — but it provides no quality assurance for manufacturing standards, purity, or safety.

Retraction Watch (retractionwatch.com) is the primary resource tracking retracted publications. MEDLINE marks retracted articles with [Retracted Publication] in the citation, and PubMed displays a prominent retraction notice. Retraction rates have increased over time (partly reflecting improved detection). Common retraction reasons: data fabrication/falsification (~60% of retractions), plagiarism (~15%), duplicate publication (~10%), and errors (~10%). For peptide compounds with limited evidence bases, a retraction of key supporting publications can fundamentally change the compound's credibility — this is particularly relevant for research compounds where the evidence base may rely on a small number of publications from a limited number of research groups. PeptideTrace notes retraction status where relevant, particularly for compounds like Dihexa where retraction history affects evidence interpretation.

Pre-specified subgroup analyses (defined in the protocol before unblinding) are more reliable than post hoc subgroup analyses. Standard subgroups in peptide drug trials include: age (<65 vs ≥65), sex, race/ethnicity, baseline disease severity (HbA1c <8.5% vs ≥8.5%, BMI categories), background therapy, and geographic region. Interaction tests assess whether the treatment effect genuinely differs across subgroups (significant interaction p-value <0.05) rather than simply testing significance within each subgroup. Important principle: even pre-specified subgroup analyses have reduced statistical power (smaller sample sizes) and increased false positive risk from multiple comparisons. Treatment guidelines generally do not vary recommendations based on subgroup analyses unless the interaction is strong, biologically plausible, and replicated.

Systematic review methodology (PRISMA guidelines): (1) formulate answerable question (PICO format — Patient, Intervention, Comparator, Outcome), (2) develop search strategy (comprehensive, reproducible search across multiple databases — PubMed/MEDLINE, Embase, Cochrane CENTRAL, Web of Science), (3) screen titles/abstracts and full texts against pre-specified inclusion/exclusion criteria (typically performed independently by two reviewers with disagreement resolution), (4) extract data using standardised forms, (5) assess risk of bias in included studies (using validated tools — Cochrane RoB 2 for RCTs, ROBINS-I for observational studies), (6) synthesise results (narrative or quantitative/meta-analysis), and (7) assess certainty of evidence (GRADE framework — rating confidence as high, moderate, low, or very low). Systematic reviews of GLP-1 RA efficacy and safety synthesise data from dozens of RCTs and provide the most reliable overall effect estimates.

Transgenic models relevant to peptide drug development include: human receptor knock-in mice (expressing the human version of a drug target receptor — necessary when the peptide has species-specific receptor binding, e.g. human GLP-1R knock-in mice for more translatable GLP-1 RA pharmacology), human FGFR3 knock-in mice (modelling achondroplasia for vosoritide development), amyloid precursor protein (APP) transgenic mice (Alzheimer's disease models for neuroprotective peptide research), and tumor-bearing transgenic models (genetically engineered mouse models that spontaneously develop tumours expressing specific receptors). Transgenic models provide more clinically relevant pharmacology data than standard animal models when significant species differences exist in drug-target interactions. The development of humanised mouse models has been a significant advance for preclinical peptide drug evaluation.

Xenograft models involve implanting human tumour cells (cell line-derived xenograft, CDX) or patient-derived tumour tissue (patient-derived xenograft, PDX) into immunodeficient mice (nude, SCID, NSG strains). CDX models use established cell lines (standardised, reproducible but may not reflect tumour heterogeneity); PDX models use fresh patient tissue (better recapitulating human tumour biology but more variable and expensive). For peptide-based cancer therapies: proteasome inhibitor (bortezomib, carfilzomib) efficacy is tested in multiple myeloma xenograft models; radiopharmaceutical peptide (Lu-177 dotatate) efficacy is tested in neuroendocrine tumour xenografts expressing somatostatin receptors. Xenograft models assess: tumour growth inhibition, survival extension, dose-response relationships, and combination therapy synergies.