Lab Research Introduction

Our lab specializes in performing AI and Machine Learning research to help agents understand and communicate with humans on various levels. To that end, most of our work involves leveraging large amounts of human behavior or communication data to learn models of communication and behavior recognition.

Research Activities

We will be running a machine learning model to predict chronological age given DNA methylation data. We are exploring different avenues such as using xgboost as an initial feature selector and also biclustering, since DNA methylation data has very many features. These will then be used to train a Deep Kernel Learning model which uses a neural network as a feature extractor before passing the final layer into the Gaussian process. This builds on the work of the Horvath Clock and the GPAge clock which are an elastic net and a Gaussian process trained to predict chronological age given DNA methylation data. This will be used to write a paper and also be submitted to an age prediction competition.

Research Projects

The main project will be a general machine learning model on the whole dataset. We will also run a subset of the data that contains only skin samples to train a separate clock.

Computational Method

We will use xgboost to get feature importance and then biclustering to select the best cpg sites. These should all be available in R. Potentially python will also be used to run a deep kernel method using PyTorch and GPyTorch

Software

R, Python

Students

Aaron Li

Browser not supported