Multi-Task Time Series Analysis applied to Drug Response Modelling

TL;DR

This paper introduces a novel multi-task learning framework for personalizing time series models, particularly in drug response prediction, leading to improved accuracy and uncertainty estimation for individual patients.

Contribution

It develops a new multi-task learning approach tailored for time series data, applicable with or without control inputs, enhancing personalization in physiological modeling.

Findings

Improved predictive accuracy over state-of-the-art models

Enhanced uncertainty estimation for individual predictions

Demonstrated on physiological drug response data

Abstract

Time series models such as dynamical systems are frequently fitted to a cohort of data, ignoring variation between individual entities such as patients. In this paper we show how these models can be personalised to an individual level while retaining statistical power, via use of multi-task learning (MTL). To our knowledge this is a novel development of MTL which applies to time series both with and without control inputs. The modelling framework is demonstrated on a physiological drug response problem which results in improved predictive accuracy and uncertainty estimation over existing state-of-the-art models.