Flávia E. Rius
flaviaerius.bsky.social
Flávia E. Rius
@flaviaerius.bsky.social
I work with Polygenic Risk Scores for Brazilians @ Mendelics
#rstats #bioinformatics
Tip for formatting text on Google Docs by Gemini *without changing content*:

If the text is in one language, use the prompt to format in another language. It worked perfectly for me!

Before that, every time I asked it to "format without changing text" it failed.

#gemini
November 11, 2025 at 1:44 PM
My talk on #positconf2025 is available on Youtube!

Check it out: youtu.be/phh8p2ImBMQ?...
Precision Medicine for All: Using Tidymodels to Validate PRS in Brazil (Flávia Rius) | posit::conf
YouTube video by Posit PBC
youtu.be
November 10, 2025 at 10:58 AM
Reposted by Flávia E. Rius
🧬 PRSs in practice (3/5): Higher-risk results = behaviour change?

A common concern is that higher-risk polygenic risk scores (PRSs) don’t motivate lifestyle behaviour change.

My @nathumbehav.nature.com paper with @mikeinouye.bsky.social addresses this 👉 rdcu.be/eNeKN

#Genomics #BehaviourChange
November 6, 2025 at 2:27 PM
So interesting!

Highest SNP heritabilities (which mean variation explained by genetics) after controlling for parent education and birthplace were around 10% for social sciences and journalism, for arts and humanities, and for natural sciences, mathematics and statistics.
November 4, 2025 at 2:40 PM
About PRSs in the clinics: people do not understand risk in general, not only the risk related to PRS.

The solution is educating health professionals to communicate PRS risk (and other risks) clearly.
🧬 PRSs in practice (1/5): Public understanding — are we underestimating the public?

A common claim: “The public won’t understand polygenic risk scores (PRSs)” — so using them in healthcare could be risky.
But is that really true? 🧵
November 3, 2025 at 5:49 PM
Reposted by Flávia E. Rius
4 word horror story for an R user:

"has a Java dependency"
October 12, 2025 at 4:29 PM
Probably late, but his week I finally understood the meaning of prompt engineering.

I used a prompt, extracted results, measured them, then adjusted both the input and prompt for better results, measured them, adjusted inputs again, measured results, and finally applied to a new dataset.
October 11, 2025 at 12:19 AM
Reposted by Flávia E. Rius
Ever wondered why analyzing RNA-seq data feels like walking through a fog with 20,000 dimensions?
Let’s talk about the curse of dimensionality in bioinformatics—and why it’s not just a math problem, it’s a biological one. 🧵
October 9, 2025 at 1:15 PM
I just want to share that I had a blast at #positconf2025! Met so many wonderful people, and gave a talk for the first time. It was awesome!
September 27, 2025 at 4:28 PM
Reposted by Flávia E. Rius
I’m new to bluesky,so.. hi!
I’m Beatriz / Bea, from Brasil, and currently I’am a post doc researcher.

I started to learn how to program in R in 2018 to use it on my research, and have been programming since then.

I have a blog where I post and share presentations etc

beamilz.com

#rstats 👋
Content – Beatriz Milz
beamilz.com
October 29, 2024 at 2:52 AM
I have just experienced the magic of rv in recovering my R packages after having to format my laptop.
Completely recommend!
github.com/A2-ai/rv

#rstats
GitHub - A2-ai/rv
Contribute to A2-ai/rv development by creating an account on GitHub.
github.com
August 24, 2025 at 3:37 PM
Every #Rstats user should read this post from @jennybryan.bsky.social‬.
Especially beginner to intermediate-level ones.
www.tidyverse.org/blog/2017/12...
Project-oriented workflow
Advice on workflows for developing R scripts. How to think about whether an action belongs in the script or elsewhere.
www.tidyverse.org
August 1, 2025 at 8:37 PM
Reposted by Flávia E. Rius
My biggest, loudest job application advice:

WHY do you want this job? This SPECIFIC job? What about it interests you? What do you think of it? What can you bring to it?

I know this sounds obvious but a generic cover letter isn't doing you favours - really. *Show me you know the job!*

1/7
July 25, 2025 at 12:09 PM
TIL you can set different values of alpha based on categories on #ggplot2.

Set 'aes(alpha = categories_col)', and then use scale_alpha_manual() to set specific transparency proportions.

I used a lower alpha in order to visualize the most important categories here.

#rstats
July 16, 2025 at 1:34 AM
Reposted by Flávia E. Rius
1/ When should you use a package vs. solving from scratch in bioinformatics?
I get asked this all the time. Here’s my approach: 👇
July 10, 2025 at 1:15 PM
TIL the parameter .before in mutate() from {dplyr}.

It adds the new column(s) created to the *beginning* of the dataframe, instead of to the end. Quite useful!

Source: #RforDataScience (2e)
(sometimes you should read stuff you think you already know)

#rstats
July 7, 2025 at 7:40 PM
Reposted by Flávia E. Rius
“The idea that taking walks, reading things unrelated to your research, and hanging out with strangers in a campus pub should be considered part of the serious process of thinking, but might well meet with skepticism in practice.”

www.pnas.org/doi/10.1073/...
The forgotten half of scientific thinking | PNAS
The forgotten half of scientific thinking
www.pnas.org
June 26, 2025 at 1:23 PM
Use "export" + the variable name and value when creating a chain of bash scripts, for it to be available in the called script too.

For example:
export LD_CORRELATION=0.2
bash run_ld_correlation_reference.sh
bash run_ld_correlation_samples.sh

#bioinformatics
June 11, 2025 at 4:10 PM
Reposted by Flávia E. Rius
I think prenatal genetic testing should become part of IVF. The cost of even a single embryo mix up is enormous. NIPT (for parental relatedness) is affordable and can be done early during pregnancy. Parents must have assurance about their child.

thenightly.com.au/australia/mo...
Monash IVF admits second major embryo mix up
Reproductive giant Monash IVF has released a statement confirming a second major mishap where a woman received the wrong embryo.
thenightly.com.au
June 10, 2025 at 12:32 PM
Quick #bioinformatics tip:

If you'll use plink files in your pipeline, whenever possible, transform your VCF into pgen or bgen ASAP!

Plink files are much faster to work with.
June 6, 2025 at 7:46 PM
I need to say that Lonely Planet is still better than chatGPT to find interesting attractions for a trip.
June 5, 2025 at 11:28 PM
jq is so useful.
June 5, 2025 at 8:11 PM
I have the impression, when reading papers from the 1990's and before, that concepts were explained more clearly than nowadays.
May 27, 2025 at 6:34 PM
I have just used #Gemini 2.5 Flash as a reviewer for my Quarto doc, and it gave me interesting recommendations.

My prompt asked for both text and #Rstats code reviews.

It returned strong points and improvement suggestions.

Useful tool especially for academics.
May 23, 2025 at 9:14 PM
Today I went to the weekly seminar of the department I do my postdoc in (FMUSP), mostly remotely.

It was great to interact with other researchers in-person. I miss this sometimes.
May 22, 2025 at 9:44 PM