James Hawley @jrhawley@scholar.social

I'm a PhD candidate in Medical Biophysics at the University of Toronto studying epigenetics and cancer. Math and physics background taking a step into bioinformatics and evolution #introductions

Thanks to everyone who has helped me get here and I wish you all the luck in your future endeavours!

It has been a privilege to work with so many talented and hard-working colleagues @pmcancercentre@twitter.com and @MBPatUofT@twitter.com, I appreciate the mentorship from my supervisor, @MatLupien@twitter.com, and am grateful for how much I have learned from my committee members @michaelhoffman@twitter.com and Dr. John Dick.

After many years, I'm delighted to say that I am finally #PhDone!

I've written a few thoughts that I'd like to share with people about the whole experience.

https://jrhawley.ca/2021/12/02/phd-lessons-learned

My PhD defense is tomorrow and I can't stop thinking about this song:

https://www.youtube.com/watch?v=bm51ihfi1p4

I'm delighted to share that my paper with @StanInScience@twitter.com and @MatLupien@twitter.com is out today in Cancer Research!

https://cancerres.aacrjournals.org/content/early/2021/10/11/0008-5472.CAN-21-2056

We use Hi-C in primary tumours to investigate how the 3D genome is altered, or not, in prostate cancer. Congrats and thanks to all of my co-authors.

For a guided breakdown of this paper that's aimed at a broader audience, here's an accompanying blog post:

https://jrhawley.ca/2021/10/12/hic-prostate-cancer

We hope you enjoy the paper and are happy to discuss this work with anyone.

This is a great video summarizing some issues with machine learning models in production.

https://www.youtube.com/watch?v=zkbPdEHEyEI

tl;dr when there are distributional shifts between training and real world data, some models can still be capable of doing what you want but implicitly learn the wrong objective. In the end, they don't do what you want and can do precisely what you don't want. This can happen even if you specify objectives accurately and perform safety checks along the way.

> I am pleased to inform you that your above-referenced manuscript has been accepted for publication in Cancer Research.

Just got confirmation that my paper got accepted today and should be published with a few weeks! I'm happy to see that out, it's been a long time coming.

For org-mode users, native citation support is here!

https://blog.tecosaur.com/tmio/2021-07-31-citations.html

This looks fantastic with a very simple syntax. I'm looking forward to using this.

Today I finally got provisional acceptance of my paper! Still some work to do, but nice that I know this paper is going to be done

I started my work/academically-oriented website years ago with no real plans for it. But I kept writing stuff for it, mostly for my own benefit. Now I've been getting emails from people saying they're reading it, have learned something from the stuff I've written. Some have even pointed out issues in some of my code and I've worked to fix it.

It's not much, but it feels good having written something people find useful. Small victories!

If you ever want a crash course in bad web design, try submitting a journal article or asking for access to data from an institution. Few things bug me more than trying to fill out these absolutely horrendous online web forms.

They don't throw a ton of data at you because they don't even need a lot of experiments.
It's just a careful, mathematically-based, and biologically-inspired analysis with clearly stated conclusions.

More of this, please.

Instead of relying on a trendy idea in the field, they slow down, develop toy models of gene regulation based on existing literature, explore how these models might function, then make the models more complex as needed.

It's so refreshing to see a paper like this one by Xiao, Hafner, and Boettiger treat the conflicting observations about TAD perturbations and enhancer-promoter contacts seriously.

https://elifesciences.org/articles/64320