CoChat

ADAM-1: An AI Reasoning and Bioinformatics Model for Alzheimer’s Disease Detection and Microbiome-Clinical Data Integration

Open access
Huang, Ziyuan, Kaur Sekhon, Vishaldeep, Sadeghian, Roozbeh, Vaida, Maria L., Jo, Cynthia, McCormick, Beth A., Ward, Doyle V., Bucci, Vanni, Haran, John P. · 2025 · IEEE Access

The DOI matches its Crossref record; 33 of the 34 references it cites were found in a registry or an index, and 1 are not checked yet.

Abstract

Alzheimer's Disease Analysis Model Generation 1 (ADAM-1) is a multi-agent reasoning large language model (LLM) framework designed to integrate and analyze multimodal data, including microbiome profiles, clinical datasets, and external knowledge bases, to enhance the understanding and classification of Alzheimer's disease (AD). By leveraging the agentic system with LLM, ADAM-1 produces insights from diverse data sources and contextualizes the findings with literature-driven evidence. A comparative evaluation with XGBoost revealed a significantly improved mean F1 score and significantly reduced variance for ADAM-1, highlighting its robustness and consistency, particularly when utilizing human biological data. Although currently tailored for binary classification tasks with two data modalities, future iterations will aim to incorporate additional data types, such as neuroimaging and peripheral biomarkers, and expand them to predict disease progression, thereby broadening ADAM-1's scalability and applicability in AD research and diagnostic applications.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/access.2025.3599857

Cites

33 of 34 references verified0 citation errors · 0 not checkable

Statistics

Figures

FIGURE 1. The Nursing Home Cohort Clinical Data Description (source: [28]). (A) Bar plots illustrate the distribution of participants based on Alzheimer’s status and the prevalence of key comorbidities. 67.16% of participants do not have Alzheimer’s disease, while 32.84% do. The most prevalent comorbidities include hypertension (HTN), and various cardiovascular diseases (CVD), with prevalence categorized by Alzheimer’s status. (B) The demographic characteristics of the cohort (N=335) encompass age distribution, with a mean age of 84.5 years and a median age of 86.0 years, alongside a pie chart depicting gender distribution, which shows a predominance of females (85.7%) in the cohort. (C) Study participation metrics include the distribution of stool samples per participant and a longitudinal sampling timeline for the top 10 participants based on the number of samples collected. The median number of samples per participant is 3. Alzheimer’s status is color-coded across all visualizations.
FIGURE 1. The Nursing Home Cohort Clinical Data Description (source: [28]). (A) Bar plots illustrate the distribution of participants based on Alzheimer’s status and the prevalence of key comorbidities. 67.16% of participants do not have Alzheimer’s disease, while 32.84% do. The most prevalent comorbidities include hypertension (HTN), and various cardiovascular diseases (CVD), with prevalence categorized by Alzheimer’s status. (B) The demographic characteristics of the cohort (N=335) encompass age distribution, with a mean age of 84.5 years and a median age of 86.0 years, alongside a pie chart depicting gender distribution, which shows a predominance of females (85.7%) in the cohort. (C) Study participation metrics include the distribution of stool samples per participant and a longitudinal sampling timeline for the top 10 participants based on the number of samples collected. The median number of samples per participant is 3. Alzheimer’s status is color-coded across all visualizations.CC BY · PubMed Central
FIGURE 2. Nursing Home Cohort Bacteria Data Description (source: [28]). (A) Prevalence vs. Abundance of Bacterial Species: Scatter plot illustrating bacterial species by their prevalence (the number of samples in which each species appears) and proportional abundance across the dataset. Red dashed lines represent the filtering criteria applied during preprocessing: species present in fewer than 5% of samples (prevalence cutoff: 17) or with a total relative abundance below 1e-4 were removed to minimize noise and focus on biologically relevant taxa. (B) Volcano Plot of Differential Abundance: log2 fold change versus –log10 (p-value) for all bacterial species. Species with significantly higher abundance in Alzheimer’s patients are shown in red; those lower in AD are shown in blue.
FIGURE 2. Nursing Home Cohort Bacteria Data Description (source: [28]). (A) Prevalence vs. Abundance of Bacterial Species: Scatter plot illustrating bacterial species by their prevalence (the number of samples in which each species appears) and proportional abundance across the dataset. Red dashed lines represent the filtering criteria applied during preprocessing: species present in fewer than 5% of samples (prevalence cutoff: 17) or with a total relative abundance below 1e-4 were removed to minimize noise and focus on biologically relevant taxa. (B) Volcano Plot of Differential Abundance: log2 fold change versus –log10 (p-value) for all bacterial species. Species with significantly higher abundance in Alzheimer’s patients are shown in red; those lower in AD are shown in blue.CC BY · PubMed Central
FIGURE 3. Workflow of the ADAM Framework. This diagram illustrates user interactions within the ADAM system. When a user submits a query along with new laboratory data, the framework employs the agentic system to refine the query and identify relevant literature evidence from a semantic search engine, integrating it with existing laboratory data. The semantic search engine collaborates with the agentic system, alongside the LLM, to reason, analyze, and interpret data, ultimately delivering precise analytical reports.
FIGURE 3. Workflow of the ADAM Framework. This diagram illustrates user interactions within the ADAM system. When a user submits a query along with new laboratory data, the framework employs the agentic system to refine the query and identify relevant literature evidence from a semantic search engine, integrating it with existing laboratory data. The semantic search engine collaborates with the agentic system, alongside the LLM, to reason, analyze, and interpret data, ultimately delivering precise analytical reports.CC BY · PubMed Central
FIGURE 4. Agentic System Workflow. The Diagram illustrates the coordinated interaction among three AI agents–Computational, Summarization, and Classification–within the agentic system. Each agent leverages Chain-of-Thought reasoning and communicates with the LLM to generate structured outputs. The system collectively produces Alzheimer’s-specific summarization and classification reports by integrating computational analysis with contextual interpretation.
FIGURE 4. Agentic System Workflow. The Diagram illustrates the coordinated interaction among three AI agents–Computational, Summarization, and Classification–within the agentic system. Each agent leverages Chain-of-Thought reasoning and communicates with the LLM to generate structured outputs. The system collectively produces Alzheimer’s-specific summarization and classification reports by integrating computational analysis with contextual interpretation.CC BY · PubMed Central
FIGURE 5. Semantic Search Engine Architecture. Semantic Search Engine. (A) The embedding model converts documents into and out of numerical vectors. (B) Vector databases work in parallel with the semantic search engine that returns the most relevant information measured by cosine similarity. Inside a Vector Database. Each publication is split into 2000-character vector chunks, with a 20 percent overlap between the prior chunk of text and the next.
FIGURE 5. Semantic Search Engine Architecture. Semantic Search Engine. (A) The embedding model converts documents into and out of numerical vectors. (B) Vector databases work in parallel with the semantic search engine that returns the most relevant information measured by cosine similarity. Inside a Vector Database. Each publication is split into 2000-character vector chunks, with a 20 percent overlap between the prior chunk of text and the next.CC BY · PubMed Central
FIGURE 6. Comparative Analysis of XGBoost and ADAM. (A) Density plots and (B) boxplots of F1 scores comparing XGBoost and the ADAM framework across multiple runs. ADAM demonstrates a higher median F1 score with reduced variance, suggesting more consistent and reliable classification performance for AD prediction.
FIGURE 6. Comparative Analysis of XGBoost and ADAM. (A) Density plots and (B) boxplots of F1 scores comparing XGBoost and the ADAM framework across multiple runs. ADAM demonstrates a higher median F1 score with reduced variance, suggesting more consistent and reliable classification performance for AD prediction.CC BY · PubMed Central
FIGURE 7. Performance Comparison of F1 Scores Between XGBoost and ADAM. Line plot comparing F1 scores of XGBoost and ADAM across 30 random seeds. Each line connects paired runs for a given seed. ADAM consistently shows improved or more stable F1 performance relative to XGBoost, highlighting its robustness and reduced variability across repeated evaluations.
FIGURE 7. Performance Comparison of F1 Scores Between XGBoost and ADAM. Line plot comparing F1 scores of XGBoost and ADAM across 30 random seeds. Each line connects paired runs for a given seed. ADAM consistently shows improved or more stable F1 performance relative to XGBoost, highlighting its robustness and reduced variability across repeated evaluations.CC BY · PubMed Central
FIGURE 8. SHAP waterfall plot for Sample FB151 (AD-positive). Malnutrition and Neglecta timonensis contributed strongly to the prediction, while the absence of protective Faecalibacterium prausnitzii further increased AD risk.
FIGURE 8. SHAP waterfall plot for Sample FB151 (AD-positive). Malnutrition and Neglecta timonensis contributed strongly to the prediction, while the absence of protective Faecalibacterium prausnitzii further increased AD risk.CC BY · PubMed Central
FIGURE 9. SHAP waterfall plot for Sample FB128 (AD-negative). The presence of Faecalibacterium prausnitzii lowered the prediction probability, counteracting other risk factors such as Neglecta timonensis and seizure medications.
FIGURE 9. SHAP waterfall plot for Sample FB128 (AD-negative). The presence of Faecalibacterium prausnitzii lowered the prediction probability, counteracting other risk factors such as Neglecta timonensis and seizure medications.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

1 more claim about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 34 references

Research works printed but not linked: 13

  1. Radford, 2018, Improving Language Understanding by Generative Pre-Training
  2. Radford, 2025, Language models are unsupervised multitask learners, OpenAI, 1, 9
  3. 2024, The Claude 3 Model Family: Opus, Sonnet, Haiku
  4. 2025, OpenAI GPT-4.5 System Card
  5. 2025, OpenAI O1 System Card
  6. Kavukcuoglu, 2025, Gemini 2.5: Our Most Intelligent AI Model
  7. 2025, Claude 3.7 Sonnet and Claude Code
  8. Jin, 2024, Demystifying large language models for medicine: A primer, arXiv:2410.18856 · arXiv
  9. 10.1126/science.ade2574 · doi:10.1126/science.ade2574 · Paper page
  10. 2024, GPT-4o-Mini: Advancing Cost-Efficient Intelligence
  11. 2024, OpenAI Platform Documentation: Embeddings
  12. 2025, Pricing
  13. Iversen, 1987, Analysis of variance, 487

Cited by

See all 14

Metadata

Authors
Huang, Z., Kaur Sekhon, V., Sadeghian, R., Vaida, M. L., Jo, C., McCormick, B. A., Ward, D. V., Bucci, V., & Haran, J. P.
Affiliations
  • University of Massachusetts Chan Medical School (US)
  • Department of Geriatric Medicine and Gerontology, Johns Hopkins University, Baltimore, MD, USA
  • Harrisburg University of Science and Technology (US)
Venue
IEEE Access
Published
2025
Volume
13
Pages
145953-145967
ISSN
2169-3536 (ISSN-L)
OpenAlex
W4413277664
Cited by
10
References
25
Citations by year
2026: 10
FWCI
6.73
Type
journal article
Language
en
Open access
Locations
  • IEEE Access (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (acceptedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University)
Funders
National Institutes of Health (R01AG067483-01)
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Alzheimer's disease research and treatments
Keywords
Alzheimer's disease, multi-agent large language models, agentic systems, multimodal data integration, microbiome data, clinical data, disease classification, XGBoost comparison, disease progression prediction

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: ADAM-1: An AI Reasoning and Bioinformatics Model for Alzheimer’s Disease Detection and Microbiome-Clinical Data Integration. https://app.cochat.ai/papers/10.1109/access.2025.3599857
The paper
Huang, Z., Kaur Sekhon, V., Sadeghian, R., Vaida, M. L., Jo, C., McCormick, B. A., Ward, D. V., Bucci, V., & Haran, J. P. (2025). ADAM-1: An AI Reasoning and Bioinformatics Model for Alzheimer’s Disease Detection and Microbiome-Clinical Data Integration. IEEE Access. https://doi.org/10.1109/access.2025.3599857

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.