Img auto Disease Learn

Learning diseases & prediction assessment

Model external medical knowledge

Automated disease learning is carried out from first principles, producing mathematical representations of diseases as a starting point in the inTrigue methodology. This is done through modelling the semantics of our target rare disease and also its symptoms and comorbidities via representation learning with large medical corpora. This semantic model is of great utility as part of predictability assessment that we conduct in each rare disease project.

Predictability Assessment

That is what customers like us to assess for them: they want to know whether or not patient finding for their disease within a given dataset is achievable. Predictability assessment determines whether this objective is achievable by any machine learning algorithm.

This requirement arises when complex algorithms are applied to real-world data which can be sparse, gappy and of poor quality, where it seems questionable whether the problem in hand can be solved.

At Volv we want to ensure a task with our customers can be achieved before they commit to a large, high value project. This allows us to preserve our reputation for successful delivery; being able to deliver on our offer is crucial for Volv.

A common approach to predictive modelling in life sciences is to leverage domain knowledge to specify features or attributes of the problem, which are assumed to be predictive of the outcome of interest, and then to build a model based on these features or attributes.

Unfortunately, this strategy often produces models which deliver poor performance.

Volv has developed a different methodology, in which the predictability of the problem is first assessed. Early analysis to determine predictors is a key way in which to assess whether or not there exist features within a data set, or range of data sets, that will allow the prediction to be successful in the first place. This knowledge can be used to inform about other factors, such as the requirement for extra data sets or for domain curation that may need to be taken into account to be able to build high performance predictive models.

Volv has a ReThink workshop methodology that lets you uncover the art of the possible, where we can cover the value of predictability assessments. We always would recommend this before embarking on any ground breaking work that can transform your business.

Learn More about Volv data science and how we make a difference
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value
Please enter a value