Abstract:
|
One route to learning how genetic changes modulate cancer outcomes is the creation of large clinical-genomic knowledge banks. We have compiled such knowledge banks from thousands of patients across several tumour types. Key observations to emerge include: Many driver mutations correlate with clinical features at first presentation, including stage and conventional risk factors for outcome; With sufficiently large knowledge banks of matched genomic and clinical data, it is possible to generate predictions of future disease course that are personally tailored to a given patient's cancer; These predictive models generally outperform current standard prognostic schemes; Personalised patient predictions require new methods for data visualisation and presentation; Knowledge banks can provide decision support for challenging therapeutic conundrums; To gain the accuracy of prediction required for supporting individual patient decisions, power calculations indicate that knowledge banks will require sample sizes of thousands; There remain statistical challenges in making "out-of-cohort" predictions from a knowledge bank to a real-world patient.
|