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

This seems like a pretty good attempt to replicate some of the mathematical methods of #SciPy in #Rust.

https://aftix.xyz/home/bacon/

Does anyone have experience trying to use a package like this for complex computational work in rust?

We performed Hi-C sequencing in frozen prostate tumours, combined our data with other genetic, transcriptomic, and epigenetic data on the same tumours, and investigated how mutations in prostate cancer rewire the genome and affect gene expression genome-wide.

We hope that these biological and technical insights will enable better characterization of patient tumour samples and give more insight into the assaults to the genetic architecture in cancer

I'm happy to share the preprint for recent work led by myself and @StanInScience@twitter.com looking at the 3D genome in prostate cancer as it goes into peer review.

"Cis-Regulatory Element Hijacking by Structural Variants Overshadows Topological Changes in Primary Prostate Cancer" is available on bioRxiv, now.

https://www.biorxiv.org/content/10.1101/2021.01.05.425333v1

I was planning on submitting a paper today. Two days ago my 11 year old motherboard crashed and corrupted my OS on my hard drive.

I have code saved on GitHub, big files and figures saved on OneDrive, and made 2 separate backups of my data.

48 hours and a new CPU + motherboard later, I have all of my data back and am up and running.
Still on track to submit today.

You know how people always say back up your data? Listen to those people

Overall, the conclusion of "detection rates are much lower than we'd like" is probably true, but is something already discussed without this paper. The numbers in this paper are tough to assess because there is little empirical evidence, and the empirical evidence they do have don't necessarily align with their predictions.

On the flip side, for smaller regions, the confidence intervals are smaller, but these don't match well with the serological estimates. Half of the regions have model predictions that don't align with the serological studies.
So there are important discrepancies between this model and independent experiments that need to be resolved.

Importantly, these discrepancies are always less than the empirical data. Meaning the model may predict fewer infected individuals than there actually are.

For large regions (Ile-de-France, Grand Est), the predictions are close to seological result, but the confidence intervals are very large. This isn't necessarily a problem, it just demonstrates that noise in the model is hard to control.

Empirical results from serological tests come from 3 studies ("EpiCoV study" [1] and "SpF study" [2], and Carrat et al. [3]) are the most important evidence for testing the model and are presented in Extended Data Figure 6.

[1] https://drees.solidarites-sante.gouv.fr/IMG/pdf/er1167.pdf
[2] https://www.medrxiv.org/content/10.1101/2020.10.20.20213116v1
[3] https://www.medrxiv.org/content/10.1101/2020.09.16.20195693v1

The findings, if true, in this paper are important.

https://www.nature.com/articles/s41586-020-03095-6

This paper suggests that between 60% and 90% of novel COVID-19 cases in France were not recorded by their healthcare system. This is a particular problem for the common strategy of test-trace-isolate that is used by many countries for handling the pandemic.

My main issue with this paper is that it's almost entirely model-based, with little empirical data or randomized experiments to support their findings.

Why does every new #single-cell paper "reveal" something?
There are thousands of articles on single-cell measurements that have "reveal" in their title.

https://pubmed.ncbi.nlm.nih.gov/?term=%22single%2Dcell%22%20%22reveals%22

This was already a meme a year or so after the first set of single-cell sequencing protocols were published. I'm surprised that it's still going on, years later, and that journals haven't come up with a better word

My PI is trying something new with the recent publications from our lab.

"Discovery Notes": a 5-10 min video explaining the major findings from a recent publication.

https://www.youtube.com/watch?v=eADV_NIWCEw&feature=emb_logo

Here's the most recent one for a publication on hematopoietic stem cells and how the structure of DNA relates to how quickly they can turn over.

Maybe someone will like this format!

I really like this.

https://tracy.posthaven.com/the-truth-about-starting-a-startup

This is a thoughtful and honest discussion of what it's like to work hard for something and how life is always going to play a role

This is a good analysis of cancer deaths in Canada, and the trends over time.

https://www.erikdrysdale.com/cancer_canstats/

Key takeaways:
1. Canada’s aging population accounts for the increase in aggregate and per capita cancer death rates
2. Age-specific cancer death rates are falling for all groups, except for the 85+ which are stable

I've been working on figures for my paper for a while now. I decided to collect some personal thoughts and tips from colleagues and others about use effective use of colour in scientific figures.

https://jrhawley.ca/2020/11/26/scientific-colours

This was a long time coming from a postdoc in my lab, but his paper is finally out on quiescent stem cells!

https://www.cell.com/cell-stem-cell/fulltext/S1934-5909(20)30539-7

In day-to-day #bioinformatics, I've needed to modify my `PATH` environment variable. This can be a bit tricky and full of pitfalls, so I made a command line tool to modify it more safely and sanely

https://github.com/jrhawley/pad-path

I've also written a blog post about it, if you'd like to read that, too

https://jrhawley.github.io/2020/11/16/pad-path

If you find this tool useful, let me know! Feedback and pull requests are definitely welcome

Feedback, questions, and boosts always welcome for what I've described

I work with #hic data in my work. One part of analyzing Hi-C data that comes up often is aggregate peak analysis. Outside of the supplementary details of a few papers, I haven't seen a lot of discussion or code on how to do it or how to do it well.

I'm hoping to change that with this blog post

https://jrhawley.github.io/2020/10/29/hic-apa