Prediction of Late Outcomes Based on Early Time Points

Participants

• Suyan Tian
• Mayte Suárez-Fariñas
• Joel Corrêa da Rosa
• Lewis Tomalin

Goal

Develop an algorithm for classification and regression that can be used for predicting late outcomes based on early time-course biomarkers.

Methods

As discussed in Zhang et al.(2013), we apply LDA-projection to convert time-course gene expressions into a time-course score that can be used as univariate input to regularization methods.

Results

The adaptation has been tested in a dataset for predicting response-to-treatment in psoriatic patients based on time-course gene expression with more than 93% accuracy.

Dissemination

This adaptation will be disseminated through a R package that can be used to select biomarkers and time points in order to predict a late outcome. We present below the pseudo-algorithm for training any classification algorithm.