Questions

My research directions are guided by certain questions I’m interested in answering or elucidating. Here is a list of some of them:

Computational Membrane Biology

1. Can transcriptomic profiling predict the nanoscale proteomic organization of the cell’s plasma membrane and/or organelle membranes? And how should lipidologic information be factored into the prediction model?

2. To what extent do the nanoscale organizations of cell/organelle membranes differ across healthy / diseased individuals for a given disease? And can we utilize this for more granular disease subtyping / disease progression modeling? E.g. for type 2 diabetes subtyping / progression.

3. Can nanoscale organization of cell/organelle membranes be used as a (functional) biomarker of aging? Is there a “membrane clock”?

4. How can the next generations of targeted drug delivery vehicles take into account the membranes’ nanoscale organization?

Automated Scientific/Technological Discovery

1. Can we design a system for identifying biological products as potent biotechnologies for a queried task? That is to say, can we provide as input something like “control cells via light” and receive an output of something like “channelrhodopsin”? See this article by Ed Boyden and Brian Y. Chow for more thoughts.

Practical Machine Learning

1. Preface: Lots of deep learning experiments are conducted by trial and error: starting with a basic neural network configuration, and model parameters/hyperparameters are adjusted until a desired performance measure is achieved. However, one can imagine that such a positive relative result (relative to other trials with other setups) may have been a fluke, and may not replicate on other similar datasets / other splittings of the data. So, my question is, can we design a measure of machine learning experiment robustness to capture an expectation on reproducibility across data/model perturbations?