Researcher

ML-Powered Biomarker Discovery — Design a Publication-Ready Study in 30 Minutes

30min vs 2-4 weeks for ML study designHealth & Medical5 min read

Key Takeaway

The Non-Tumor ML Research Planner skill generates complete machine learning study designs for non-cancer diseases — from GEO dataset selection to diagnostic model building to immune infiltration analysis. One prompt produces a study framework that would take a research team weeks to plan, with 4 workload tiers from 2-week sprint to 14-week publication.

The Problem

You're a researcher who wants to use machine learning to discover diagnostic biomarkers for a non-cancer disease — say, rheumatoid arthritis, Alzheimer's, or diabetic nephropathy. The challenge:

Which GEO datasets are relevant and high-quality?
Which ML algorithms are appropriate for your sample size?
How do you handle feature selection with thousands of genes?
How do you validate your model properly (not just overfit)?
What's the right way to integrate immune infiltration analysis?
How do you design this for actual publication, not just a class project?

Designing a rigorous ML study takes 2-4 weeks of methodological planning. Junior researchers make predictable mistakes: they don't correct for batch effects, they overfit with aggressive feature selection, they use accuracy instead of AUC on imbalanced datasets, and they skip external validation.

The Solution

The Non-Tumor ML Research Planner generates complete study designs following 5 validated patterns, with explicit methodology choices, statistical thresholds, and failure contingencies at every step.

The Process

View details

You: Design a machine learning study to identify diagnostic
biomarkers for early-stage Alzheimer's disease using
publicly available transcriptomic data.

The agent generates a structured study design:

View details

## ML Biomarker Discovery — Early Alzheimer's Disease

### STUDY PATTERN: DEG-to-Diagnostic
(Differential Expression → Feature Selection → ML Model)

### SPECIFIC AIMS
Aim 1: Identify differentially expressed genes between
  early AD patients and cognitively normal controls using
  GEO transcriptomic datasets (DISCOVERY)
Aim 2: Build and validate a diagnostic classifier using
  ensemble ML with cross-validation and external test set
  (MODEL)
Aim 3: Characterize immune cell infiltration patterns
  associated with early AD using computational deconvolution
  (MECHANISM)
Aim 4: Identify hub genes through PPI network analysis and
  validate clinical relevance through survival/progression
  correlation (TRANSLATIONAL)

### DATASET SELECTION
Training datasets (merged):
- GSE63060: Blood transcriptome, 145 AD + 104 controls ✅
- GSE63061: Blood transcriptome, 139 AD + 135 controls ✅
- Combined n=523 (sufficient for ML)

External validation:
- GSE97760: Independent cohort, 36 AD + 36 controls
- GSE122063: Brain tissue (different tissue = stronger
  validation)

Quality checks:
☐ Platform compatibility (all Illumina HumanHT-12)
☐ Batch effect correction (ComBat or limma::removeBatchEffect)
☐ Sample quality (boxplots, PCA for outliers)
☐ Clinical covariate balance (age, sex, APOE status)

### ANALYTICAL PIPELINE (14 Steps)

Step 1: Data preprocessing
  Tool: R/Bioconductor (GEOquery, limma)
  Normalization: quantile normalization
  Batch correction: ComBat (sva package)
  Threshold: Remove probes with detection p>0.05 in
  >50% of samples

Step 2: Differential expression analysis
  Tool: limma
  Model: ~disease_status + age + sex + APOE
  Thresholds: |logFC| > 0.5, adj.p < 0.05
  Expected: 200-800 DEGs

Step 3: Feature selection (3-layer approach)
  Layer 1: DEG filtering (Step 2 output)
  Layer 2: LASSO regression (glmnet, lambda.min)
  Layer 3: Random Forest variable importance (top 50)
  Intersection: genes passing ≥2 of 3 methods
  Expected: 15-40 candidate biomarkers

Step 4: ML model building
  Algorithms: Random Forest, SVM (radial kernel),
  XGBoost, LASSO logistic regression
  Validation: 10-fold cross-validation, repeated 100×
  Metric: AUC-ROC (primary), sensitivity, specificity
  Threshold: AUC > 0.80 for clinical utility

Step 5: Ensemble model
  Method: Weighted average of top 2 performing algorithms
  Weight: proportional to CV-AUC

Step 6: External validation
  Apply model to GSE97760 and GSE122063
  Report: AUC, sensitivity, specificity, calibration plot
  Decision: AUC > 0.75 on external set = proceed to paper

Step 7: CIBERSORT immune deconvolution
  Tool: CIBERSORT (LM22 signature matrix)
  Analysis: Compare immune cell fractions AD vs control
  Correlation: hub genes vs immune cell proportions

[Steps 8-14: PPI network, GO/KEGG enrichment, hub gene
identification, clinical correlation, figure preparation,
manuscript outline]

### WORKLOAD TIERS

| Tier | Timeline | Figures | Target Journal |
|------|----------|---------|---------------|
| Lite | 2-4 weeks | 5-6 | Brief communication |
| Standard | 4-8 weeks | 8-10 | Specialty journal |
| Advanced | 8-14 weeks | 12-16 | High-impact specialty |
| Publication+ | 14+ weeks | 16-20 | High-impact general |

### COMMON PITFALLS (pre-empted)
⚠️ Overfitting: 10-fold CV + external validation prevents this
⚠️ Batch effects: ComBat + PCA visualization before analysis
⚠️ Imbalanced classes: Use AUC not accuracy, apply SMOTE if needed
⚠️ Multiple testing: Benjamini-Hochberg correction for all DEG
⚠️ Data leakage: Feature selection INSIDE CV folds, not before

The Results

Metric	Manual Study Design	AI Agent
Time to complete design	2-4 weeks	30 minutes
Methodological completeness	Variable (experience-dependent)	14 steps, all documented
Pitfall prevention	Learned from reviewer rejection	Pre-empted in design
Dataset recommendations	Hours of GEO browsing	Curated in minutes
Workload estimation	Optimistic guesses	4 calibrated tiers
Reviewer-ready	After 2-3 revisions	First draft quality

Setup on MrChief

yamlShow code

skills:
  - non-tumor-ml-research-planner
  - medical-research-toolkit
  - pubmed

Start free on Mr.Chief →

machine-learningbiomarker-discoveryGEO-datasetstranscriptomicsdiagnostics

Related case studies

Patient

Your AI Doctor on Call — Symptom Assessment Without the 3-Week Wait

An AI agent with the Doctor skill provides instant symptom assessment, emergency recognition, medication safety checks, and first aid guidance — accessible 24/7 through Telegram. It doesn't diagnose, but it tells you whether to call 911, go to urgent care, or schedule an appointment.

30s vs 4h ER wait for symptom triage7 min read

Athlete

Your AI Gym Coach — Progressive Overload Tracking That Actually Works

The Gym skill logs every workout, tracks PRs, enforces progressive overload rules (+2.5kg or +1-2 reps per session), adapts for injuries, programs deload weeks, and warns when you're training the same muscle group too soon. Like having a coach in your pocket for $0/month.

Plateau detection after 3 stagnant sessions5 min read

Athlete

Apple Health Meets AI — Ask Questions About Your Own Fitness Data

The Apple Health Skill connects your agent to your Apple Health data via the Transition app. Ask natural language questions — "How has my resting heart rate changed this month?" — and get answers based on YOUR actual metrics, not generic advice.

Natural language queries on personal fitness data4 min read

Want results like these?

Start free with your own AI team. No credit card required.

Start Free →Browse agents