Emy Yue Hu's thesis focuses on graph machine learning for network biology, for which she wanted to use a type of ML method called graph neural networks. Last summer, she went to Mila, Montréal’s machine learning institute and the partner of the Helmholtz International Lab Causal Cell Dynamics (CCD) to learn this processing technique, with the support of Prof. Jian Tang’s lab. Now, she tells us more about her time in the centre!

 

How did this Scientific Stay impact your PhD project?

I started an exciting project, fueling my curiosity and giving me more motivation for the PhD itself. I had been wanting to learn more about using graph neural networks for link prediction in biological graphs for a while. Upon wrapping up an old project, I finally had the chance to even go to Mila for the pursuit of that goal.

We predict the function of a gene based on its underlying biological regulation and found performance better than state-of-the-art alternative methods. It is the first time to specifically model the interactions between genes/proteins for the prediction of gene function and I am seeing a lot of potential of exploring these biological knowledge graphs.
 

What are the things you like the most about your stay at Mila?

Scientifically: Mila is an AI hub with around 800 researchers. There are many research directions with specialized reading groups for in depth discussions, such as one about graph representation learning. The authors of papers are invited for a talk and one benefits greatly to learn from them about the emerging topics in that field.

Socially: It’s easy to get in touch with researchers from different groups due to Mila’s open working spaces. The big open kitchen is the main place to meet people during tea breaks or for lunch. Everyone was open and friendly, so that even on the first day, I met so many people and was already whisked away for lunch and in the evening for the Jazz festival downtown. In addition, there are many activities to join such as dancing, Frisbee, and baking club.

What will you take back from your Scientific Stay there?

I learned so much on the scientific side, but also worked on my communication and collaboration skills, especially how to bridge computational biology and theoretical AI. I learned to be more confident in my knowledge gaps and dare to ask more questions and ask for help. I value the collaboration with Zhaocheng and Sophie from Prof. Tang’s group, as we complement each other nicely.

It was a great opportunity giving a joint talk at the CCD symposium in front of renowned researchers such as Fabian Theis, Yoshua Bengio, Bernhard Schölkopf and Doina Precup.

Further, when Bundeskanzler Olaf Scholz and Premiere minister Justin Trudeau came for a visit at Mila during my stay, I -among others- got the chance to say hello and discuss our work a little bit. That was fun :)

For my personal development, it was a great opportunity to meet people from different cultural backgrounds, getting to know them and forging deep friendships.

How will the project you started there continue since you're back?

While I was there, we had in-person meetings which have now shifted to the virtual realm. They are productive, even though being able to write and draw on white board makes discussions easier. Interestingly, only after I got back, we decided to submit an extended abstract about our project to the Machine Learning conference for Computational Biology organized by MIT/Stanford ML. It was rather short notice. We had many meetings at that time and split the work well. Finally, we received the notification that our abstract was accepted for an oral presentation.

For the future, we will continue with our collaboration online. But soon, Sophie will be visiting Helmholtz, so that we can have part of the discussion in person again. I am looking forward to hosting her in our group, as this is another great opportunity for both her and me/my group to learn from one another.

What would you recommend to other PhD students who want to visit Mila?

Definitely go for it ;)

More on a general note: It wouldn’t hurt to have some project outline laid out, as to get into the project fast. But also, be aware that still things can go wrong. Due to some troubles with the data, our original plan was not possible. Mid-way through the research stay, we had to come up with something new from scratch. Wanting to make the best of my time there, I felt pressure to devise something fast. Luckily, after some trial and error, there were some promising ideas we could explore. Things don’t always play out as one hopes, so take some uncertainty into account. A more relaxed stance would have helped me not to panic too much.

Have regular meetings with your collaborators. Be curious and join some reading groups. Go out and have fun with your peers.

Any final words?

I am grateful for this opportunity to broaden my horizon and learn from the best. Thanking Annalisa Marsico and Jian Tang, as well as CCD and the Joachim Herz Foundation!