nference - Collaborations for Pharma
Our scientists use the nference platform and data in collaboration with your teams to tackle your most important questions and challenges. nference is creating the largest labeled EMR dataset in healthcare. EMR data contains longitudinal real-world, “deep data” rich in clinical phenotypes and outcomes spanning across therapeutic areas, but exists in in largely semi-structured and unstructured form. The nference platform powers the tranformation of the raw data into a labeled EMR dataset.
We leverage our deep biomedical expertise to transform semi-structured data (e.g. lab tests, procedures) into structured data through technology-enabled data harmonization.
We build upon state-of-the-art neural networks and Natural Language Processing (NLP) methods to transform unstructured clinical text (e.g. clinical notes) into structured patient features.
Triangulation across data and state of the art algorithms power collaborations
Connecting structured, harmonized EMR data with publicly available literature, molecular and real world datasets and applying our powerful proprietary algorithms allows us to ask and answer challenging questions
Triangulation With Public And Proprietary Data
Integrating additional data sources with labeled, curated and harmonized EMR data can strengthen the insights derived from downstream data analysis. This data includes biomedical literature (ex. PubMed), clinical trials (ex. clinicaltrials.gov) and molecular data (ex. Gene Expression Omnibus).
The labeled, curated and harmonized EMR data serves as a starting point that enables training state-of-the-art AI algorithms. We develop these algorithms as solutions to some of the major challenges in healthcare, including early detection and diagnosis of disease, identification of biomarkers for disease progression and more.
Work with our scientists to tackle hard problems in Pharma R&D, Medical Affairs and Strategy
Our scientists will use the nference platform and data in collaboration with your teams to tackle the questions and challenges most important to you
Identify candidate targets with significant evidence for potential therapeutic benefit for a selected indication
Evaluate likelihood of drug success pre-launch and identify opportunities for commercialized assets
Clinical Trial Design
Build complex patient cohorts using real world EMR data to optimize clinial study criteria and study design. Read our Publication
Real World Outcomes
Track patient-level outcomes following specific treatments or interventions
Identify biomarkers or correlates of disease progression or treatment outcomes using our clinico-genomic data which spans therapeutic areas
Identify indicators of disease or disease progression to support earlier, targeted treatment or intervention
Leverage the longitudinality of our dataset and power of our technology to identify early predictors of disease
Identify drug side effects and understand the populations most susceptible to them based on real world use
Identify disease symptoms and understand the populations most susceptible to them
Get disease area specific insight on the end-to-end patient journey, leveraging longitudinal patient data
Existing EMR Data
Decades of rich but unprocessed data
Electronic medical records (EMRs) contain a tremendous amount of information about human health, but is largely unstructured. The amount of data from Mayo Clinic alone - 500M+ clinical notes, 4B+ images, 1B+ lab results and more - pose a challenge and an opportunity
Pathology digitization and molecular sequencing
We digitize tens of millions of archived pathology slides using high-throughput state-of-the-art scanning capabilities in partnership with Pramana, and we work with sequencing partners to provide clinical-grade whole exome and transcriptome sequencing for patients seen at our partner centers. These data creation processes enrich the real world data currently existing in EMRs.
Best-in-class algorithms and ‘Data Under Glass’
Patient and data privacy are at the core of everything we and our partners do. Our algorithms for deidentification, including for challenging data modalities such as unstructured (free-text) data, have been certified as best-in-class. Our “data under glass” approach ensures that the data, even after deidentification, always remains at the center.
Augmented Curation · Harmonization
Making healthcare data computable
The majority of EMR data exists in semi-structured or completely unstructured forms. A critical part of enabling artificial intelligence applications downstream is transforming this data into a structured labeled data. We leverage state-of-the-art technology coupled with deep biomedical expertise to transform semi-structured data (harmonization) and unstructured data (augmented curation), resulting in the largest labeled dataset in healthcare.
Neural Networks · Triangulation
Creating AI-enabled solutions for healthcare
We use the labeled data on its own to train state-of-the-art neural networks that enable next-generation diagnostics and precision medicine. We also triangulate the labeled data with other datasets such as public literature, molecular and real world data to create solutions and support scientific discovery.