Nicola Romanò @nicolaromano@qoto.org
Senior lecturer at the Zhejiang-Edinburgh Joint Institute (ZJE) and Edinburgh University.
Undergraduate Programme Director, Biomedical Informatics at ZJE.
I teach #imageanalysis & #dataanalysis with #RStats & #python. I study #heterogeneity in #pituitary (and other) cells.
I'm also very interested in #reproducibility and #openscience.
Joined Oct 2022
@janeadams I disagree with that (to a point). Maybe it doesn't count towards your graduation requirements, but doing a PhD is first of all about growing as an independent researcher, developing your critical thinking and finding your way in (work) life.
Requirements are there and you should of course complete them, but I suspect those other things will be helpful in the future (of course if you spend 90% of your time on things outside your PhD project, maybe you should change your PhD topic!)
@michaele @amckinstry Those machines also have a phone number you can call to pay if you don't want to use an app.
Nicola Romanò
boosted
Which scientific publishers/journals are worst affected by fraudulent or dubious research papers, and which have done least to clean up their portfolio?
A science-integrity startup called Argos says it has answers.
There are quite a few integrity tools now that look for red flags in papers, but this is the first to go public with what it's finding across journals and publishers.
Here's my exclusive look at their figures.
https://www.nature.com/articles/d41586-024-03427-w
Nicola Romanò
boosted
“tinytable is a small but powerful R package to draw beautiful tables in a variety of formats: HTML, LaTeX, Word, PDF, PNG, Markdown, and Typst. The user interface is minimalist and easy to learn, while giving users access to powerful frameworks to create endlessly customizable tables.” - @vincentab
@terence There's quite a lot of data available in the wild, for example those from this paper https://pmc.ncbi.nlm.nih.gov/articles/PMC10088239/ which you can find here https://zenodo.org/records/6025935
Nicola Romanò
boosted
[New paper]: Two subtle problems with over-representation analysis.
ORA is a type of enrichment analysis that analyses over-represented functional categories in gene lists. These tools have accumulated ~190k citations, but they have subtly different behaviours. Here we unpack the differences and investigate two subtle problems in some implementations, which may have negatively impacted those 190k research papers.
https://doi.org/10.1093/bioadv/vbae159
#genomics #bioinformatics
Nicola Romanò
boosted
I’m a software developer with a bunch of industry experience. I’m also a comp sci professor, and whenever a CS alum working in industry comes to talk to the students, I always like to ask, “What do you wish you’d taken more of in college?”
Almost without exception, they answer, “Writing.”
One of them said, “I do more writing at Google now than I did when I was in college.”
I am therefore begging, begging you to listen to @stephstephking: https://mstdn.social/@stephstephking/113336270193370876
I've created a little schematic on basic Git/GitHub usage.
Feel free to reuse! (CC-BY-NC-4.0)
Nicola Romanò
boosted
Six tips for going public with your lab’s software: https://www.nature.com/articles/d41586-024-03344-y
1) make time for maintenance
2) simplify installation
3) add a GUI or good CLI
4) good documentation
5) use github/git
6) automated testing
Any other tips people have? #SoftwareEngineering #opensource #openscience
Nicola Romanò
boosted
@computingnature
These cover a lot of common problems I have with scientific code, one minor one to add is "build more, smaller things" - splitting up a very large monolithic code base into several smaller pieces can work with all the above tips to make the software much more useful over time. Being able to re-use pieces between projects without needing to make future projects direct dependents of humongous prior packages is a huge deal for labs that might make many tools.
@mccarthymg @jimbob @djnavarro Yes, I teach a lot of biologists who are completely scared of programming. A student the other day told me they get anxious whenever they get a red line in the console in R Studio (in that particular instance, we were actually installing a package and all these red lines continue popping out although there was no error... grrrr).
A great thing to do when dealing with non-technical audiences, is to make coding mistakes on purpose so that you can analyse the error messages that you get with them. You demistify the error message and they start to be a bit more comfortable with the whole programming thing. And then you get them to read the manual, or Google the error code, or hell even use chatgpt if they're onto that. But the important thing is you have them engage with their code.
Another thing that I often see in my field is that people often have complex data, and they want to analyse it and get a results out of it. They don't care about programming details, or how the underneath algorithms work, and in a sense I understand them. However, you need a good balance between the two if you are to trust your own results.
Nicola Romanò
boosted
#academicchatter I read a published paper on the other day and noticed a panel in the figure is clearly duplicated but by mistake, I knew this because the authors uploaded the raw data as table and the table has a complete set of numbers that doesn't match the figure panel. Importantly the correct data didnot change the conclusion. This
#openscience approach either enforced by publishers or volunteered by authors maintained the trust I have with the finding they reported.
Nicola Romanò
boosted
"My paper was proved wrong. After a sleepless night, here’s what I did next"
column by @oaggimenez describing how to gracefully react when mistakes are discovered in ones published papers.
Nicola Romanò
boosted
@zaunkoenig @jonny sorry, but, if you are talking about scientific research, it's absolutely not true that any complex model uses NNs nowadays. You can verify this yourself by checking out any reputable scientific journal, for example those of the APS (the Physical Review ones).
Nicola Romanò
boosted
@jimbob @mccarthymg @djnavarro But... you need to have a well written manual and you need to be able to understand it... which is not a given (and that's why we need to also train people on his to do that)
Nicola Romanò
boosted
to celebrate selling out the first print run of How Integers and Floats work, we're giving away 500 PDF copies of the zine!
use code BUYONEGIVEONE at checkout to get a copy for free. (no need to enter your real address)
As usual this works with the honour system, this code is for you if $12 USD is a lot of money for you!
Nicola Romanò
boosted
Wow it transports 20 people. ooooh I have an idea, what if we connected a bunch of these together... in a long chain and put them on tracks to make them more energy efficient (and to meet power needs) then ran them along the most popular transportation corridors in major cities! They could even go in tunnels in places like NYC to reduce traffic!
Golly Elon is on to something this time!
Nicola Romanò
boosted
#statstab #198 Bayesian mixed effects (aka multi-level) ordinal regression models with {brms}
Thoughts: Useful tutorial also for frequentists, as it covers checking multiple links at once in {ordinal}.
#ordinal #brms #clmm #probit #cloglog #r #cauchit
https://kevinstadler.github.io/notes/bayesian-ordinal-regression-with-random-effects-using-brms/
@eamon Why opening them at all? I just mass select and delete!
Luckily, my work email is only maybe 40% junk?
Senior lecturer at the Zhejiang-Edinburgh Joint Institute (ZJE) and Edinburgh University.
Undergraduate Programme Director, Biomedical Informatics at ZJE.
I teach #imageanalysis & #dataanalysis with #RStats & #python. I study #heterogeneity in #pituitary (and other) cells.
I'm also very interested in #reproducibility and #openscience.
Joined Oct 2022
