Disease-Specific Studies

Validated across therapeutic areas.

High-level results from KolateAI work samples and pilot studies conducted with design partners. Therapy and product identifiers are anonymized where required by partner confidentiality. Full detail is available under NDA.

4 therapeutic areas
14 validation studies
90%+ accuracy
Immunology, Dermatology

Therapy Response Prediction: TCS & Dupilumab in Atopic Dermatitis

Despite multiple treatment options for AD, topical corticosteroids (TCS) and systemic biologics like dupilumab, clinicians still rely on trial-and-error prescribing. KolateAI's multi-modal ingestion agent processed ~1,400 real-world dermatology records (EHR & clinical notes, verified by a KolateAI dermatologist) to build patient-level response models for both first-line TCS and dupilumab, identifying treatment-resistant patients before they were prescribed and surfacing the clinical factors that most drive response.

97.9%TCS Accuracy
95.1%Dupilumab Accuracy
1,400Patients
Applications: trial enrichment · in-silico therapy comparison · point-of-care screening
Oncology

Synthetic Head-to-Head: CAR T-Cell Therapy in Large B-Cell Lymphoma

In partnership with a CAR T-cell therapy developer, KolateAI trained response and safety prediction models on a real-world 2023–24 dataset of ~260 patients on the partner's CD19-directed therapy and ~260 patients on a previously approved comparator. Running the partner therapy's model on the comparator's actual patients produced an "apples-to-apples" synthetic head-to-head, with no new prospective trial required.

84%Response Model Accuracy
71% vs 65%Predicted vs Actual CR
520Patients
Applications: in-silico therapy comparison · quick, inexpensive market differentiation evidence
Metabolic Disease

Population Screening: Type 1, LADA & Type 2 Diabetes

Distinguishing Classic T1D, Latent Autoimmune Diabetes in Adults (LADA), and T2D from routine data alone is difficult, and LADA is frequently misdiagnosed as T2D, delaying appropriate insulin therapy. Using NHANES 1999–2023 multi-cycle survey data (~7,772 diabetes patients, 57,204 including the full screening population), KolateAI trained and compared 10 model families per experiment to flag at-risk patients for each subtype.

99.8%T2D Model AUC
98.3%T1D Balanced Accuracy
57,204Patients
Applications: population screening · early identification for targeted intervention
Metabolic Disease

Metabolic Syndrome Risk & Simulated Klotho Modulation

In partnership with a biotech developing a Klotho-based therapy, KolateAI ran a retrospective NHANES cohort study to identify the key predictors of Metabolic Syndrome and simulate the patient-level impact of modulating serum Klotho via digital twins, intended as a trial-enrollment screening tool and, eventually, a point-of-care application.

92%Model Accuracy
93%AUC-ROC
4,855Patients
Applications: trial enrollment screening · digital-twin therapy simulation
Working in a Different Therapeutic Area?

The platform is extensible to new indications.

Talk to us about a proof-of-concept for your program.

Contact us for a POC