Inspect an author's raw data, methods, and reproducibility across their publications.
Press Enter ↵ to find
Fuel Your Discoveries
"The most beautiful thing we can experience is the mysterious. It is the source of all true art and science."
- Albert Einstein
Quick Explanation
Copied
Jingjing Guo (author disambiguation caution)
OpenAlex returns multiple “Jingjing Guo” author records with different ORCIDs, so any metric attribution depends on correct disambiguation. One “top_author” snippet shown in the input is labeled “Li Guo” (not “Jingjing Guo”).
What looks strong: the works summarized in the input include (i) multi-modal protein representation learning with structure+sequence, (ii) plant molecular signaling with mechanistic protein degradation evidence, (iii) pathogen-host specialization with population genetics, and (iv) multiple additional in silico / biochemical / materials papers—suggesting cross-domain competence.
Long Explanation
BGPT Author Review — Jingjing Guo
Date: 2026-03-31 Scope: scientific-strength critique using only the evidence explicitly provided in the prompt: OpenAlex author metadata snapshot + a set of paper-level raw-data summaries (DOIs listed in the prompt).
1) Critical disambiguation & metric reliability
OpenAlex ambiguity: the provided OpenAlex snippet contains multiple “Jingjing Guo” matches with different ORCIDs and different citation/work counts; however, the snippet’s “top_author” block is labeled “Li Guo” and includes a full counts_by_year array that does not necessarily correspond to “Jingjing Guo.”
Implication: any citation/H-index claims must be treated as conditional on the correct ORCID mapping; for this review I therefore (a) report paper-level evidence where DOIs are provided and (b) treat author-level bibliometrics as uncertain unless the prompt explicitly ties metrics to the Jingjing Guo ORCID in question.
2) Evidence set used (papers explicitly provided)
The prompt provides paper summaries with DOIs (example set): BridgeNet (protein ML) ; VNS–gut–butyrate–vagus–brain epilepsy axis ; PUB12/13 and ABI1 degradation in ABA signaling ; and plant-pathogen specialization in rice blast fungus (Magnaporthe oryzae) ; plus additional computational chemistry/materials/in vitro items listed in the prompt (e.g., High-PepBinder, starch derivatives, graphene oxide methods, influenza genome/virulence).
3) Visual: bibliometric time series (from the provided ‘top_author’ block)
Important: the prompt’s only explicit yearly array is attached to an author block labeled “Li Guo”, not explicitly to “Jingjing Guo.” I therefore present it as informational only about the shown “top_author” time profile, not definitive for Jingjing Guo.
4) Paper-level scientific strength: what the author’s work (in the provided set) appears to do well vs. what remains weak
Strength pattern #1 — Mechanistic anchoring (wet-lab mechanistic logic when present)
The plant ABA-pathway paper explicitly describes protein–protein interaction, ubiquitination, proteasome dependence, genetic rescue, and RNA-seq overlap—i.e., it attempts to triangulate mechanism rather than only show correlation.
Strength pattern #2 — Population-genetics + experimental infection design (when host specialization is studied)
The Magnaporthe oryzae study (eLife) links field sampling/genotyping and structured isolates to cross-inoculation phenotypes and effector/Pi diagnostics, providing a more causally oriented framework than “sequence-only clustering.”
Strength pattern #3 — Multimodal / latent-space integration in computational biology (but generalization demands caution)
BridgeNet’s description indicates a structured attempt to bridge sequence and structure modalities and uses ablations across multiple tasks. However, the summary also notes dataset dependence (UniRef 50 representation biases; PDB structural coverage bias) and evaluation breadth that may not capture all protein families.
Strength pattern #4 — Causal chain testing attempts in neuro-immuno-metabolic axis work (but small n & cross-domain translation)
The VNS–gut–butyrate–vagus–brain summary includes longitudinal human sampling plus convergent mouse interventions (FMT, vagotomy, antibiotics enabling FMT, AKK colonization, butyrate dosing). Still, it reports a small clinical cohort (n=11) and a translational gap (KA model vs human epilepsy).
Recurring weakness / blind-spot in the provided set:
Several computational or translational papers in the prompt rely on in silico surrogates (e.g., AlphaFold3/Rosetta/MD scores for binding; metagenomics/biomarker associations without fully randomized causal trials at the human level). Such studies can be scientifically valuable, but their ground-truth validity and transferability need careful validation beyond the reported surrogate metrics.
5) Visual: extracted performance metrics (from the prompt’s paper summaries)
These are only the numeric endpoints explicitly provided in the prompt’s extracted “results_and_conclusions” blocks; they are not re-derived from the full papers here.
Source endpoints: BridgeNet EC Fmax 86.6%, GO-MF 66.7%, coenzyme AUROC 0.989, peptide toxicity AUROC 98.2 as stated in the prompt’s BridgeNet summary ; High-PepBinder affinity classifier AUC 0.869 as stated in prompt .
6) Scientific quality verdict (evidence-weighted)
Overall (based on the provided evidence set only): the author’s represented work spans multiple methodological styles (mechanistic wet-lab signaling, evolutionary host-pathogen specialization, and computational protein/peptide modeling). Across those categories, the prompt evidence suggests competence in experimental design and mechanistic argumentation when wet-lab is involved (e.g., ABI1 degradation framework) and nontrivial causal testing attempts in translation-oriented axis work (VNS/FMT/vagotomy/colonization) .
But: the set also contains computational studies whose claims are necessarily mediated by surrogate scoring (AlphaFold/Rosetta/MD; latent embedding metrics) and whose generalization to truly unseen biology can be limited by training data and structural coverage.
Confidence in this verdict:moderate, because (1) author disambiguation in the OpenAlex snippet is not guaranteed and (2) the review uses prompt-provided extracted summaries rather than directly re-mining full texts here.
What would most likely falsify / revise this view?
Correcting disambiguation reveals the cited papers are not actually attributable to the intended “Jingjing Guo” ORCID.
For the computational works, independent evaluation shows that the reported improvements do not replicate on truly out-of-distribution proteins/peptides beyond the benchmark regime summarized in the prompt.
For translation-like human studies, larger cohorts with better confounder control show no consistent microbiota/butyrate/vagal dependency pattern despite plausible mechanistic models.
7) Suggested next BGPT actions (optional)
Feedback:
Updated: March 31, 2026
BGPT Author Review
Scientific Quality
70%
Based on the provided evidence summaries, the author appears capable across multiple biological subdomains and can present mechanistic experimentation (e.g., protein degradation signaling) and more structured inference approaches (e.g., population-genetics-informed host specialization). However, the evidence set is heterogeneous and includes computational/surrogate-based studies where generalization and wet-lab validation are not shown in the prompt. Additionally, the OpenAlex snippet contains disambiguation risk (a “top_author” block labeled “Li Guo” is not clearly tied to “Jingjing Guo”), reducing confidence in attribution of bibliometrics to the intended person.
Communication Quality
60%
The prompt’s summaries are generally structured (methods/results/limitations), but communication quality cannot be judged from author summaries alone; mechanistic and modeling works are presented clearly in the extracted text, yet there is not enough direct information about writing style, clarity of claims, or calibration of uncertainty in the underlying papers.
Author Novelty
70%
Novelty appears moderate-to-high for the provided computational framework descriptions (multimodal latent bridging; diffusion-based affinity-aware peptide design) and for biological mechanistic framing (receptor-dependent degradation module in ABA signaling; effector repertoire shaping host specialization). Still, without full-text context and with reliance on extracted summaries, novelty is inferred rather than directly validated.
Scientific Rigor
70%
Rigor looks reasonably strong where mechanistic triage is described (interaction + ubiquitination + proteasome dependence + genetic rescue + RNA-seq overlap; or population genetics + cross-inoculation + effector diagnostics). Rigor is comparatively weaker for studies relying primarily on surrogate metrics (computational protein/peptide binding scores) and for translational biomarker/correlation work with small human n and model-system dependence.
We'll email you the results when your analysis is finished.
Hypothesis Graveyard
The apparent improvements in binding/toxicity prediction from multimodal models are primarily artifacts of benchmark label leakage rather than true generalization—unlikely but cannot be ruled out without strict leakage audits across datasets.
The microbiota–butyrate–vagus mechanism in epilepsy is entirely epiphenomenal and disappears under larger replication cohorts with stronger confounding control—plausible if the current human sample is not representative, but the prompt’s mouse causal chain attempts argue against a purely epiphenomenal account.
Fuel Your Discoveries
Quick Explanation
Long Explanation
Top Data Sources
Ask a Follow-Up
Key Insight
Keep Exploring
Hypothesis Graveyard
Potential Experiments
Discussion