GPMelt: an overview

As said in the introduction, GPMelt is not just a ‘black box’.

Our goal is to give you a clear understanding of how GPMelt works, helping you specify the model that best fits your data and easily understand the key parameters.

Ready to dive in? Let’s take a tour of GPMelt’s inner workings!

1 Hierarchical models

We begin by explaining the principle of hierarchy and how this is used in GPMelt (see here).

This will help us understand how we can translate an experimental protocol in a hierarchy (see here).

We continue by describing how these hierarchical models can be linked to the machine learning principle of multi-task learning (see here).

2 Model specification and estimation

After this, we begin by introducing Gaussian process (GP) regression and next dive a bit further into it.

We continue by explaining the GP parameters (here) and how they are estimated via Type II MLE. We also discuss how the model complexity can be adapted here.

Finally, we discuss why the observation should be scaled, and which scaling to choose (here).

3 Statistical framework

Next, we explain how differentially melting proteins are captured in GPMelt (see here).

4 A Nextflow pipeline

Taken together, we have the GPMelt model!

Next step is to understand how to use the Nextflow implementation, and this is the purpose of the Tutorial :)