BGPT: Author Review: Yiran Wei

Explore by Goal

Quick Explanation Copied

Author scientific strength (evidence-based)

From the provided records, Yiran Wei’s strongest demonstrated contributions (biomed) align with human cohort inference (cardiac autonomic metrics → dementia risk) and translational microbiome biomarker work (stool signatures in autoimmune hepatitis), with generally careful statistical modeling but typical observational/16S limits.

Long Explanation

Author Review: Yiran Wei

Date: April 22, 2026 • Evidence scope: ONLY the studies explicitly provided in your prompt’s extracted datasets.

1) Scientific citation metrics (from prompt-provided numbers)

h-index = 3, total citations = 28, paper count = 8 (as provided in your prompt’s “Author Citation Information”).
These metrics are not validated against a canonical database in this conversation; treat them as provisional.
Your prompt also includes an OpenAlex disambiguation list with multiple “Yiran Wei” identities and much higher counts for other matches, implying name/identity conflation is possible. Without an unambiguous ORCID-to-work mapping, cross-database metric comparisons are uncertain.

2) Evidence-grounded assessment of scientific work (2 provided datasets)

A. Cardiac HR response/recovery (UK Biobank → dementia)

Study: “Heart rate response and recovery during exercise and dementia risk: a prospective UK biobank study” in Scientific Reports (2025-11-03).

Key evidence originates from: https://doi.org/10.1038/s41598-22299-2.

B. Gut microbiome dysbiosis (AIH steroid-naïve → signatures)

Study: “Alterations of gut microbiome in autoimmune hepatitis” in Gut (2019-06-14).

Key evidence originates from: https://doi.org/10.1136/gutjnl-2018-317836.

3) Visual evidence: effect sizes & discriminative performance

4) What the provided work suggests about scientific capability

Strength signals

Statistical care in observational design: the UK Biobank extraction reports Cox regression plus competing-risk methods and mixed-effects models for cognitive trajectories, aligning with standard approaches to time-to-event and longitudinal outcomes in cohort studies.
Attempted external validation: the AIH microbiome work reports a discovery/validation design with independent cohorts and AUC metrics in both sets.
Focus on biologically plausible axes (as framed by the studies): autonomic cardiovascular dynamics ↔ neurovascular/cognitive aging; and gut–liver axis dysbiosis ↔ immune-mediated liver inflammation. Note: plausibility is not proof; it only motivates hypothesis generation.

5) Critical limits & skeptical checkpoints (what could mislead)

Observational inference risks (HRR → dementia)

Outcome ascertainment bias: dementia identification via ICD-10 codes in hospitalization records can miss milder/subclinical cases and may vary by healthcare access. The extraction flags this limitation.
Residual confounding: even with covariates, unmeasured factors (e.g., systemic inflammation, subclinical disease burden, neuroimaging markers) can create spurious associations. The extraction notes potential residual confounding.
Multiple testing / effect size modestness: the extracted HRs are near 0.9 per SD, which is plausible for risk biomarkers but can also reflect small systematic biases. The extraction does not provide a full multiplicity report here, so uncertainty remains.

Biomarker risks (AIH microbiome signatures)

16S-only functional inference: PICRUSt-style pathway enrichment is model-based and depends on reference genomes; it can be directionally wrong if taxa-to-function mappings differ by strain/background. The extraction explicitly positions functional inference as predicted (not measured).
Confounding by medications/diet: steroid-naïve is a strength, but antibiotics/UDCA exposure and other unmeasured factors can shape stool composition, producing signatures that track treatment exposure or co-morbidities rather than causality. The extraction notes medication/diet/environment confounding risk.
Biomarker overfitting risk: AUC ~0.78–0.81 is encouraging, but without external (multi-center, multi-ethnicity, different labs) replication details, performance transportability is uncertain. The extraction does include a validation cohort, but not broader generalization.

6) Overall scientific judgment (confidence-labeled)

Strength-of-evidence synthesis (based strictly on provided extracts)

Moderate evidence of analytical competence in both epidemiology (time-to-event + competing risks) and microbiome biomarker modeling with validation. Confidence: moderate (limited to only 2 provided studies).
Biological interpretation should be treated as hypothesis-generating rather than mechanistically resolved: observational design and predicted functional inference (16S) limit causal claims. Confidence: high given the stated limitations.
Possible identity ambiguity across the “Yiran Wei” records in OpenAlex indicates that attributing aggregate author metrics (h-index/citations) to the same individual could be incorrect. Confidence: moderate based on your provided disambiguation list.

7) How this review could be falsified / changed

If additional provided works reveal frequent causal experimental designs (not just association) or strong multi-center replication, then the scientific rigor/impact scores would likely increase.
If the HRR effects or microbiome AUCs fail in replication cohorts (different ethnicities, sites, assays), then “evidence strength” should be downgraded.
If authorship is misattributed due to name ambiguity (multiple “Yiran Wei” individuals), the true publication record for the target person may differ substantially.

Feedback:

Updated: April 22, 2026