The Second Brain
Microbiome-Driven Early Detection
Machine learning infrastructure for early detection of gut conditions through microbiome analysis. We analyze the microbiome data hospitals already collect and predict gut conditions 3-6 months before traditional diagnostics catch them.
Patient samples validated with 83% accuracy in predicting condition onset
Earlier detection than traditional diagnostics, enabling preventive interventions
Annual healthcare savings potential through early detection and prevention
We're catching disease too late
Right now, 60 million Americans are walking around with gut conditions developing inside them. Their bodies are sending warning signals. The data exists. But we're not looking at it.
By the time symptoms get bad enough for diagnosis, the damage is often irreversible. We treat advanced disease when we could have prevented it. We perform surgeries when dietary changes might have worked. We prescribe immunosuppressants when early interventions could have stopped inflammation before it cascaded.
4-6 yrs
Average wait between first symptoms and IBD diagnosis
90%
Five-year survival for stage I colorectal cancer
14%
Five-year survival for stage IV colorectal cancer
$136B
Annual US spending on gut condition treatment
This isn't because doctors aren't trying. It's because they lack the tools to see what's coming.
Your gut is already telling us what's wrong
Scientists call the gut microbiome the "second brain" for good reason. The 100 trillion bacteria in your digestive system don't just break down food. They regulate your immune system, produce neurotransmitters, influence metabolism, and communicate with every organ in your body.
When things start going wrong, your microbiome changes first. 3-6 months before you feel sick. Months before traditional tests show anything abnormal. Research shows microbiome composition shifts precede IBD flares by an average of 4.2 months. Colorectal cancer signatures appear in stool microbiome data 6-12 months before tumors are detectable.
Hospitals already collect this data. They sequence over 800,000 stool samples annually. Then the data sits in a folder somewhere, unused. A single sample generates data on 1,000+ bacterial species, 500+ metabolic pathways, and 50,000+ genes. The patterns are too complex for humans to process.
Per Sample
1,000+
Bacterial species
500+
Metabolic pathways
50,000+
Genes cataloged
We built the tool that should already exist
The Second Brain is machine learning infrastructure for early detection. We analyze the microbiome data hospitals already collect and predict gut conditions 3-6 months before traditional diagnostics catch them.
We've validated this on 31,000+ patient samples. Our models achieve 83% accuracy in predicting condition onset. Not by looking for obvious signals doctors could spot themselves. By finding the subtle compositional shifts, the trajectory patterns, the complex interactions that only become visible when you can process thousands of data points simultaneously.
Our models don't just say "high risk." They show which specific bacteria and metabolic pathways are changing. They give doctors actionable insights: elevated Bacteroides fragilis, reduced butyrate production, inflammatory pathway activation. Real biological mechanisms they can address with targeted interventions.
Catching conditions 3-6 months earlier reduces treatment costs by 40-60%. For a single IBD patient, that's $15,000-30,000 in avoided immunosuppressant costs. For colorectal cancer, early detection means the difference between a $50,000 polypectomy and a $200,000+ surgical oncology treatment.
Research Methodology
Our approach combines cutting-edge machine learning algorithms with deep domain expertise in microbiome science. We process massive datasets containing information on bacterial species composition, metabolic pathway activity, and functional gene expression to identify early warning signals that precede disease onset.
The models are trained on longitudinal patient data, allowing us to track how microbiome composition changes over time and identify the specific patterns that indicate developing gut conditions. This temporal analysis is crucial for early detection, as it captures the dynamic nature of microbiome shifts that occur months before symptoms appear.
Machine Learning Models
Advanced algorithms for pattern recognition in high-dimensional microbiome data
Biomarker Discovery
Identification of bacterial ratios, metabolic disruptions, and gene expression changes
Clinical Validation
Validated on 31,000+ real patient samples across multiple institutions
Key Findings
Our research has identified several critical biomarkers and patterns that serve as early indicators of gut condition development. These include specific bacterial ratios, metabolic pathway disruptions, and functional gene expression changes that consistently appear 3-6 months before traditional diagnostic methods detect disease.
The validation across 31,000+ patient samples demonstrates the robustness and clinical applicability of our approach. With 83% accuracy in predicting condition onset, our models represent a significant advancement in preventive healthcare for gut-related diseases.
By identifying at-risk patients months before symptoms appear, clinicians can implement targeted interventions that may prevent or significantly delay disease progression. This shift from reactive treatment to proactive prevention represents a fundamental change in how we approach gut health.
Prevention should be the default
The current system is backwards. We wait for people to get sick, then try to fix them. We know prevention works better and costs less, but we don't have systems that enable it.
Preventive Care
$500-2K
per patient annually
Every $1 spent on prevention saves $3-6 in treatment costs
Late-Stage Treatment
$50-200K+
per patient for advanced disease
Yet only 3% of healthcare spending goes to prevention
What if your annual checkup included microbiome screening? What if your doctor could see inflammatory bowel disease developing six months out and suggest dietary changes before you ever felt symptoms? 5,400+ hospitals in the US have gastroenterology departments. They already collect microbiome samples from 800,000+ patients. The infrastructure exists. The data exists. We just need to connect them.
This is bigger than gut health
Your microbiome influences everything. Metabolic disorders. Autoimmune diseases. Neurological conditions. Even mental health. The gut-brain axis is real, and we're just starting to understand it.
78%
Type 2 diabetes risk prediction accuracy from microbiome data
0.71
Correlation coefficient between microbiome composition and depression severity
5-10 yrs
Parkinson's onset detectable in microbiome before symptoms
If we can demonstrate that microbiome analysis enables early detection for gut conditions, we prove the concept for using it across medicine. Your "second brain" becomes an early warning system for your entire body.
If we catch just 20% of gut conditions 3 months earlier, we prevent 180 million patient-months of unnecessary disease progression annually in the US alone. That's 15 million years of combined suffering prevented every single year.
We're scientists who got tired of waiting
We founded The Second Brain because we saw the gap between what research shows is possible and what's actually deployed in clinics. We're researchers at Yee Collins Research Group. We've published the papers showing microbiome-based early detection works. We've built the models. We've validated them on 31,000+ patient samples across multiple institutions.
Our mentor and friend John Capodilupo dedicated his career to improving early detection of gut conditions. He showed us how much needless suffering happens because we catch things too late. He challenged us to build something that could actually change that.
Founded by Brandon Yee, Wilson Collins, and Kundana Kommini
Research Directors, Yee Collins Research Group
Building on the mission of our mentor and friend, John Capodilupo: comprehensive detection of early indicators of gut conditions
Published Papers
The Second Brain: Diffusion Models Learn to Generate the Human Microbiome
Novel diffusion model architecture for generating synthetic human microbiome profiles, enabling data augmentation for rare gut conditions and early detection model training.
Brandon Yee, Wilson Collins, Kundana Kommini, Maximilian Rutkowski
Status
Patent Pending
Application
Microbiome Generation
Framework
Diffusion Models
Catch it before it catches you.
If you're a hospital, gastroenterologist, researcher, or patient who thinks medicine should predict problems before they become serious—we want to work with you.