Forecasting the timings of large trials is critical for all parts of the life science ecosystem. By combining granular data on trial locations, disease prevalence and end points, we are able to accurately predict read out dates for interim and full analysis for a number of Phase 3 COVID-19 vaccine trials and also infer accurate efficacy estimates from the timings. We are building out these forecasts in a number of areas and enhancing them with more historical data sets.
Many lives and the world economy depend on the rapid scale up of COVID-19 vaccine production. Through mapping manufacturing sites and supply chain dynamics, analysing historic production rates and capacity, Airfinity generates accurate predictions on vaccine production and realistic supply forecasts.
Combining production forecasts with virus variants of concern, vaccine hesitancy, efficacy and effectiveness assumptions as well as variants of concern prevalence, Airfinity models supply and demand dynamics for vaccines and therapeutics across 140 countries.
Candidates that have/are expected to report positive late phase randomized trial results that have not yet received regulatory approval. The Priority 1* grouping is arrived at by considering several metrics (including but not limited to: clinical progress and trial parameters, funding and regulatory status), updated in real-time with expert opinion oversight.
Our data science team has developed a machine learning model that accurately predicts 1-2 week infection rates at regional and country specific levels. We are working on combining this model with the world leading COVID-19 infection forecasts developed by the epidemiology team Imperial College London to improve our ability to predict longer term infections and enhance our pandemic risk alerts.
Airfinity uses efficacy and safety data from trials in real-time to calculate the health economics of a new innovation. Combining this with pricing and deal data, we build evidence-based bottom-up market share and pricing forecasts.
Social media has become a key source of scientific information but eliminating the noise is difficult. We have built a sentiment and semantic classifier to analyse large scale social media data and derive results on overall sentiment on a particular drug.