Mayo Clinic has more than 150 years of unsurpassed expertise in medicine and an impeccable reputation for patient care, both in excellence and integrity. Aligned values and vision to transform healthcare through innovation brought Mayo Clinic and nference together. As a strategic and trusted Mayo Clinic Clinical Data Analytics Platform partner, nference will uphold the highest standards of excellence in innovation while maintaining privacy and putting the needs of the patient first.
The Clinical Data Analytics Platform is the first venture announced under the overall Mayo Clinic Platform. Clinical Data Analytics Platform will utilize advanced analytics on all of Mayo Clinic’s de-identified data to advance medicine and improve the health of patients. Mayo Clinic selected nference as its Clinical Data Analytics Platform partner to accelerate drug discovery and development across the biopharmaceutical ecosystem. nference will focus on identifying targets and biomarkers for new drugs, optimal matching of patients with therapeutic regimen, and real world data and real world evidence applications, such as label expansion, post-marketing surveillance and drug purposing. nference will be the strategic partner for analytic services to the biopharmaceutical industry.
De-identified patient health records are rich with the information needed – phenotypes, co-morbidities, outcomes - to improve the way we serve patients with new medicines, better diagnoses, etc. However, this health information that captures the underlying disease biology is often present in an unstructured form (clinical notes, discharge summaries, pathology reports, pathology images, radiology images, etc.) that was originally meant for an expert (physician/physician scientist) to interpret. To enable interpretation and analytics of de-identified patient records at scale, the unstructured information must be first structured. Few other initiatives and companies are currently tackling that problem by using manual curation, but this has been a major bottleneck that has limited their scope and impact. The nference technology makes unstructured knowledge computable and it allows the seamless triangulation with various structured databases (such as vitals, lab tests, ICD10 codes, genetic variants) that enable precision phenotyping at an unprecedented scale.
One of the major strengths of Mayo Clinic is its large repository of samples - biospecimens, pathology slides... - from patients who have consented to partake in research. nference and Mayo plan to conduct large-scale characterization of those biospecimens using a wide range of assays such as whole exome sequencing, single cell RNA seq, CyTOF... as well as digitization of Mayo's pathology slides. While there are other major ongoing clinico-genomic efforts taking place worldwide (such as UK Biobank), we believe nference and Mayo could generate the largest and deepest de-identified clinico-genomic databases by linking molecularly characterized Mayo's samples to precision phenotyping driven by automated curation of EMR data.
Mayo Clinic has 150 years of unsurpassed expertise in medicine and an impeccable reputation for patient care, both in excellence and integrity. As a strategic and trusted partner with aligned values, nference will uphold the highest standards of excellence in innovation while maintaining privacy and putting the needs of the patient first.
Mayo Clinic brings the Clinical Data Analytics Platform, Mayo Clinic’s first venture under the Mayo Clinic Platform. nference makes biomedical knowledge computable by synthesizing information from de-identified unstructured and structured clinical datasets. The toolsets built by nference enable researchers at Mayo and biopharma industry to leverage the power of its Augmented Intelligence platform on the de-identified data. Additionally, Mayo Clinic is a strategic investor in nference, contributing to the nference Series-B round, and becoming a substantial equity holder of nference.
The focus of this partnership is to unlock the value of big data for medical advancement in an ethical and privacy preserving way that can significantly benefit patients across the entire gamut of patient journey – from the discovery of new life-saving targeted therapeutics, to the advancement of new clinical research frameworks that significantly de-risk the expensive clinical development and operations processes to make future medicines more available to patients. Mayo Clinic and nference seek to be the gold standard for privacy preserving innovation and research, and to standardize Mayo Clinic’s data analytics practices across the institution. Through these combined efforts, Mayo and nference will put the value delivered to patients first and foremost across our partnership.
Along with the shared culture and values between Mayo Clinic and nference, Mayo Clinic’s equity investment provided significant accountability for nference, allowing Mayo Clinic to have much deeper insight into nference.
De-identified patient health data are rich with the information needed to improve the way we serve patients with new medicines, better diagnoses, more robust outcomes, etc. Through the Clinical Data Analytics Platform nference can access insights from the data while maintaining patient anonymity. Health information that captures the underlying disease biology is often present in an unstructured form (clinical notes, discharge summaries, pathology reports, pathology images, radiology images, etc.) that was originally meant for an expert (physician/physician scientist) to interpret. Such unstructured information are rich in biological context and therapeutic outcomes. However, they are not directly computable from a research perspective. The nference technology makes unstructured knowledge computable and enables seamless triangulation with various structured databases (such as vitals, lab tests, ICD10 codes, genetic variants). By establishing concordance or discordance of insights computed from differentiated knowledge bases at scale, our platform in turn enables scientists and physicians to answer deeper questions on disease biology, longitudinal progression of treatment outcomes, and therapeutic/diagnostic options to best serve patients. By capturing nearly 150 years of institutional knowledge, the platform has the unprecedented opportunity to help physicians and patients match therapeutic regimen to patient constitutions, significantly building on all the fantastic work to decode human genotype-phenotype relationships – right from the advent of the Human Genome Project, to recent efforts such as the UK BioBank. Examining Real World Evidence (RWE) across large cohorts of patients with similar and related conditions will be a powerful source for improving the lives of patients.
As simple as the concept is, researchers have been unable to glean real world outcomes from Electronic Health Records (EHR) in an effective way to date. The major reason for this shortcoming is that most of the valuable information in medical records is largely unstructured information. Few other initiatives and companies are currently working on structuring that unstructured EHR information, but those efforts have mostly relied on manual curation. While those efforts have shown the promise of structuring unstructured EHR information, the manual curation required in those efforts has been a major bottleneck that has limited their scope and impact. nference is advancing the capability to automate the curation of unstructured EHR information. This automated curation of EHR data together with the continuously-evolving nferX software platform – that triangulates insights from the world’s biomedical knowledge - can result in incredible advances in patient care such as early diagnosis of disease, finding patients for clinical trials to research new therapeutic solutions, and finding new drug targets to develop future medicines. Furthermore, previous attempts to garner insights from health records have been focused on specific therapeutic areas, but patient journeys across multiple physician specialties (captured by the institutional knowledge of Mayo as the leading academic medical center) provide a more genuine reflection of systems biomedicine (i.e. the inter-connected biological processes that drive disease progression and influence therapeutic outcomes). For instance, the same biochemical signaling pathways within cells are often shared by diverse disease pathologies ranging from cancers and autoimmune conditions, to neurological and cardiovascular conditions. Holistic inference of disease progression and therapeutic outcomes thus necessitates the synthesis of institutional health records at a comprehensive scale.
No. Using a federated architecture model, Mayo will maintain the de-identified patient data in a Cloud platform in Mayo’s span of control. Clinical Data Analytics Platform servers operate in that Cloud framework and exposes a privacy preserving Application Programming Interface (API) through which nference servers provide the synthesized information. Researchers access the synthesized information through the nference Cloud platform, and thus never directly access any de-identified patient level data.
The beauty of our Federated architecture is that the institution (Healthcare provider, Academic Medical Center…) remains the custodian of the patient data, while still enabling the training of AI/ML models and detection of aggregate insights. Federated Architecture enables multiple participants to build AI/ML (artificial intelligence/machine learning) models without sharing datasets, thus addressing critical issues such as data privacy, security, and access rights to heterogeneous sources of data.
In addition to removing Protected Health Information (PHI) according to the 18 categories defined under HIPAA (HIPAA Safe Harbor), the De-identification process removes other identifiers (such as care provider’s names, pharmacies, locations, etc.) to sufficiently mitigate re-identification risk. Thus, we are exceeding the HIPAA gold standards. The De-identification software and the corresponding processes are certified by a third-party statistician under the HIPAA Expert Determination Model (colloquially referred to as the “Statistical Standard”).
Because the data being utilized for Clinical Data Analytics Platform has been stripped of any identifying information and because there is no impact on a patient’s care, de-identified data can be used to advance innovation and research without specific consent. This is different from a clinical trial in which medical care is provided and identifiable information is required, both of which require consent.
The multiyear strategic partnership with Google is two-fold:
Mayo will securely store it’s data within the private Mayo Clinic Cloud, which is housed within and protected by the Google Cloud;
Mayo and Google will also engage in joint research and innovation projects that will combine Mayo’s clinical expertise with Google’s talented data engineers to advance solutions to health care’s biggest challenges. nference is the strategic analytics partner for working with the Biopharmaceutical industry within Mayo’s Clinical Data Analytics Platform initative. nference is empowered to conduct analysis on the de-identified data sets to glean insights from those data for several therapeutic applications as well as conducting formal research collaborations with Mayo Clinic.
Since the research will be on a cloud, it will be done in multiple locations. However, only authorized and trusted personnel will be conducting research, and no data will leave Mayo Clinic. Any access will be governed through a secure connection to the data hosted on Mayo Clinic’s servers using the federated learning approach.
nference is working towards partnerships with other academic medical centers and health systems.