Raeline Valbuena (@rcvalbuena) header
Raeline Valbuena (@rcvalbuena) profile picture

PhD candidate studying context-specific gene regulation @StanfordMed | @StanfordBioX fellow 🇵🇭🇧🇷 | Buying coffee and doing my best to be kind

California, USA
Joined November 2020
Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)Photo by Raeline Valbuena (@rcvalbuena)
I’m amped to share our new work to decrypt transcriptional repressors, now on bioRxiv: biorxiv.org/content/10.1101/… We paired deep mutational scans with machine learning to predict and even tune repressor activity. (1/8)
4
27
1
93
13,874
Incredibly excited that our paper linking single molecule states of TF binding to gene expression using quantitative thermodynamic models is out in Nature today. An amazing collaboration with the Bintu Lab. Congrats to Ben, Michaela, and Julia! nature.com/articles/s41586-0…
11
104
8
508
48,864
Defended my PhD dissertation on halloween, and the verdict is unanimous - thank you everyone for your support, spirit, and cheers. 💌
2
31
1,633
UPDATE: Our research and my artwork are featured on cover of Nature Immunology’s October 2024 Issue! See the feature and read our paper here: nature.com/ni/. So proud to have been a part of this amazing team @KarmiWarmii Olivia Laveroni @BrookeHowittMD @LivnatJerby
4
30
1,436
Raeline Valbuena retweeted
The paradox of poetry and information theory
'Paradoxes in Probability Theory and Mathematical Statistics' - Gábor J. Székely (1984, PDF): archive.org/details/paradoxe… Having a lot of fun with this one so far.
67
1,020
101
5,541
633,983
Lastly, another key takeaway from our work is conceptual: we HAD to consider our computational goals when designing our experiments, and biological motivations when designing our models. Interdisciplinary cross-talk could be the foundation for more collaborations to come. (7/8)
1
1
6
532
This was such cool and ultimately rewarding experience, with @akshat_ai especially, and the whole team. (8.1/8)
1
5
326
Thank you all for showing me how to be an authentic and curious scientist, and for making this work possible - @JoshTycko, @RoshniAPatel, @PeterSuzuki8, Kaitlyn Spees, Aradhana, Sophia Arana, Peter Du, @RoshniAPatel, @BintuLab, @anshulkundaje, and @MCBassik. (8.2/8)
6
316
We also used our data to functionally annotate small linear interacting motifs (SLiMs) in disordered domains: these are critical for repressor activity in the context of our screens, and give us insight into how disorder effectors function in a cell. (5/8)
1
3
326
We then used TENet to tune the activity of repressors through directed evolution. To the best of our knowledge, this includes some of the first successful attempts to alter the regulatory functions of intrinsically disordered protein domains. (6/8)
1
2
311
Then, we developed TENet (Transcriptional Effector Network) to integrate sequence, structure, and biochemical features of repressors. After training on our data, our model can accurately predict repressor activity of holdout sets, even in one-shot contexts. (6/8)
1
3
302
First we mined these data for relevant functional insights. We found thousands of coding variants with measured gain or loss of function. For example, known pathogenic variants and variants of unknown significance the TWIST1 HLH domain show significant loss of function. (4/8)
1
3
312
So we’d need really fine-grained and suitably rich data to power anything useful — we designed deep mutational scans of over 50 repressor domains. Then we measured the activities of these hundreds-of-thousands of protein variants in human cells. (3/8)
1
2
325
Repressor domains are super diverse because transcriptional repression happens in many different ways. But it has been difficult to develop accurate functional models for what can be (lovingly) considered an eclectic (and often disordered) class of proteins. (2/8)
1
3
412
Raeline Valbuena retweeted
Great excerpt. Unfortunately, we’ve made the job of a scientist, above all else, to keep the job of being a scientist — even if that path becomes increasingly unrecognizable from the work of science. While many earnestly care about expanding knowledge or ushering in the next generation of scientists, we’ll often will do literally whatever box-checking, pyramid-climbing, ego-stroking exercises are perceived to keep the doors open to clutch on to the dregs of what looks like science that remain. Good news is it’s actually in our hands but it’s a prisoners dilemma and, when it comes to science and publishing, we’re in our own way to improve it.
"The majority of science published isn't consumed by its intended audience...instead most scientists just skim new titles...We’re running around publishing for the sake of [it]...exactly the kind of environment where fraud not only exists but abounds" arstechnica.com/science/2024…
1
12
58
12,093
Raeline Valbuena retweeted
How the project happens –– How we write the paper
17
205
25
1,572
131,696
Raeline Valbuena retweeted
They ATAC, they genome HAC, and baseball SMAC on this Twilight THROWBAC. Belated Happy Halloween and Feliz Día de Los Muertos from the Bassik lab 🧛🏻‍♀️🧛🏻🧢🍎
1
1
1
17
3,082