— Data, Models and Insights —
Advanced Modeling and Analysis Services
Why Modeling Matters
-
In modern clinical trials, digital biomarkers offer the promise of richer, more sensitive insights — but realizing this value requires more than raw data. It demands expert processing, advanced modeling, and scientific rigor.
-
Our full-service modeling and analytics solutions help sponsors unlock the full potential of digital data to accelerate development, improve trial design, and strengthen regulatory submissions.
Our Services
Scalable
Data
Processing
We Developed Scalable, Cloud-Native Infrastructure for High-Throughput Sensor and Wearable Data Processing
-
Cloud-native pipelines handle terabytes of time-series data from wearables and sensors
-
Flexible architecture supports standard and custom algorithms across study designs
-
Outputs are structured, analysis-ready datasets for modeling and regulatory use
Statistical
and Machine
Learning Modeling
Our expert team designs and executes statistical modeling strategies tailored to study objectives:
-
​Development of comprehensive Statistical Analysis Plans (SAPs)
-
Longitudinal modeling of disease progression
-
Assessment of test-retest reliability and other aspects of clinical validity
-
Machine learning model development for feature selection, risk scoring, and endpoint optimization
-
Power analyses to inform study design and optimize sample size​
Scientific Consulting
and Reporting
We partner closely with client teams to translate modeling results into strategic insights
-
Support for regulatory-facing documentation
-
​Data visualization and interactive result exploration
-
Guidance on endpoint selection, adaptive design strategies, and predictive modeling
-
Ongoing collaboration to maximize commercial impact of digital biomarkers
Client Project Impact
Parkinson’s Risk Prediction
In partnership with the Michael J. Fox Foundation and Verily, we supported an initiative to detect early Parkinson’s risk using smartwatch-derived features
-
The dataset included 32 TB of smartwatch data -
(150,000 files from 350 participants).
-
Our team developed a composite machine learning predictor that effectively stratified individuals into high- and low-risk groups.
-
The resulting digital risk index predicted clinical test outcomes and supported identification of individuals at elevated risk for targeted intervention.
-
Our scaled compute allowed 10 - 20x faster delivery of extracted insights
Data Sources
Extracted Physiological Features
Activity
Gait
Vital Signs
Sleep
Integrated Machine Learning Model
Development of Risk Index
Risk is Associated with Clinical Severity
ALS Progression
A leading biopharmaceutical company sought to determine whether digital biomarkers of motor function could detect disease progression more rapidly and reliably than traditional PROs.
-
The dataset comprised ~1.5TB of tri-axial accelerometry data (~30,000 files from 450+ participants).
-
Custom algorithms were integrated in to our cloud-native pipeline to extract gait, upper limb mobility, postural transitions, and daily activity measures.
-
The resulting measures demonstrated excellent reliability and sensitivity, enabling the sponsor to identify superior biomarkers and inform future digital strategy.
-
Our scaled compute allowed 10 - 20x faster delivery of extracted insights.
Data Sources
Extracted Physiological Features
Upper Limb Mobility
Gait
Postural Transitions
Activity
Progression Analysis of Digital Measures
Reliable Metrics with Faster Detectable Change
PD Progression
In partnership with a top 10 pharmaceutical company, we deployed our neuroscience toolkit to support digital biomarker validation in Parkinson’s disease.
-
Our team successfully accommodated mid-study protocol amendments, reprocessing data with updated algorithms to ensure scientific rigor and consistency.
-
We processed 54 GB of accelerometry data (50,000 files from 100+ participants).
-
Our scaled compute allowed 10 - 20x faster delivery of extracted insights.
Data Sources
Data Processing Infrastructure
Raw Data
Controller
Status Data
Containers
Output Data
Extracted Physiological Features
Gait
Tremor
Activity
Dexterity
Models of Disease
Built on Our Digital Measures
Model | Generalized capability | Other use cases |
|---|---|---|
PD On/Off | Symptomatic levodopa dose response curve | Also applies to drugs for other neuro, pain, hypertension, ADHD, addictions etc. |
PD risk score | Prognostic biomarker using known precursors to diagnosis | Other combinations of measures in in other indications may leverage same approach |
PD Progression | Amplifying progression signal in slow variable progression | Palsies, cardiovascular disease, other neuro like HD, MS |
Ambulatory function | Functional capacity at home in indications which reduce capacity | All Cardiovascular, HPP, Asthma, Knee OA |
Neuro EEG | Prognostic biomarker using known precursors to diagnosis | EEG power spectrum analysis for neurologic changes before presentation, cardiac events prior to ischemia, signs before fever |
ALS Progression | Isolating progression rate in variable population | Dementias, PSP, diseases without progression models e.g. lysosomal disorders |
SCD Pain Crisis | Objective measure detection of events | Asthma, COPD, migraine, epilepsy, ataxias, most autoimmune disorders |
​Accelerated development timelines through
earlier, more sensitive disease signal detection​
-
​Our scaled compute allowed 10 - 20x faster delivery of extracted insights, accelerating subsequent clinical study design.
Progression of digital measures more sensitive
than traditional PROs​
-
​In the ALS progression tracking project, digital measures showed a significant increase in monthly change over time relative to the standard PRO
Data-driven protocol optimization, enabling
adaptive trial designs​
-
​In simulated analyses comparing at-home versus in-clinic study designs, denser and more precise digital measures resulted in 68% fewer patients required per trial arm (Lavine et al., 2024).
​​Regulatory-grade outputs that support confident submissions and strategic trial decisions
