Pangaea provides a machine learning based software product to its customers from the biopharmaceutical and healthcare industry for faster identification of patient cohorts based on phenotypes (clinical characteristics and symptoms) from electronic health records (EHRs) and unstructured doctors’ notes.
This is critical for detecting patients at risk of diseases, finding genes linked to a phenotype in the context of drug or biomarker discovery, recruiting patients for clinical trials and real world evidence (RWE) studies.
Find new patients across diseases in an EHR dataset
50 times faster and more accurate than NLP, indexing, semantic searches and keyword matching
Scalable to 5000 diseases
Higher Return on Investment for our customers who spend to acquire EHR data
Recent results from our work have shown that by using our machine learning based algorithms we can automatically label EHRs based on medical notes so that finding patients based on specific phenotypes is easier, quicker and much more accurate.
We have applied our algorithms to a dataset of 52,722 EHRs and were able to label them with HPO (Human Phenotype Ontology) terms in 40 mins with more than 90% accuracy. This is at least 50 times faster and 30% more accurate than current natural language processing (NLP) tools and keyword search approaches. Additionally, we were also able to discern HPO terms for 5,000 EHRs, which were not detected using keyword searches.
Our technology is based on the founders’ work over the last 20 years in industry and at Imperial College London. Pangaea is advised by renowned scientific and technical experts from the biopharmaceutical domain and Stanford University. The company is supported by teams in London, San Francisco and China.
We are looking to collaborate with scientists and clinicians from life sciences who are interested in applying our algorithms to the EHR datasets they are building or accessing through partners. Get in touch at email@example.com or send us a message using the contact form below.