@yosephbarash.bsky.social
Opinions - own. Lab RNA+AI+Genetics @upenn.edu
www.biociphers.org
www.biociphers.org
Finally, we have new models, more results and more analysis brewing in all those areas so stay tuned! 😀
November 24, 2025 at 2:32 AM
Finally, we have new models, more results and more analysis brewing in all those areas so stay tuned! 😀
#3 David Wang led MAJIQTL method development with validations by Peter Choi lab (CHOP/UPenn) and joint students Kevin Yang and Benjamin Wales-McGrath. This project started many years ago with the late and great Casey Brown to whom we dedicated this work 💔
November 24, 2025 at 2:32 AM
#3 David Wang led MAJIQTL method development with validations by Peter Choi lab (CHOP/UPenn) and joint students Kevin Yang and Benjamin Wales-McGrath. This project started many years ago with the late and great Casey Brown to whom we dedicated this work 💔
#2: Matthew Gazzara led the APA/DDX55 paper, a joint effort with co-mentor Kristen Lynch
November 24, 2025 at 2:32 AM
#2: Matthew Gazzara led the APA/DDX55 paper, a joint effort with co-mentor Kristen Lynch
#1 Farica Zhuang led G4mer with Danielle Gutman performing validations, Dan Dominguez lab (UNC) structure validation and Kate Nathanson (UPenn) helped with the Breast cancer analysis.
November 24, 2025 at 2:32 AM
#1 Farica Zhuang led G4mer with Danielle Gutman performing validations, Dan Dominguez lab (UNC) structure validation and Kate Nathanson (UPenn) helped with the Breast cancer analysis.
As I noted, each paper is a different comp approach, and a different element of RNA processing! I just find this so exciting we can do this kind of science in the lab 🙂 This is possible because (a) lab members did terrific job 👏💪(b) great collaborators 🙌😊. Specifically:
November 24, 2025 at 2:32 AM
As I noted, each paper is a different comp approach, and a different element of RNA processing! I just find this so exciting we can do this kind of science in the lab 🙂 This is possible because (a) lab members did terrific job 👏💪(b) great collaborators 🙌😊. Specifically:
we published blog posts that give an overview of what's new here: biociphers.wordpress.com/2025/01/20/a... biociphers.wordpress.com/2025/01/28/a...
A Deep Dive into sQTL Modeling – Part 2
In my previous post, I described how, in our recent sQTL preprint, we broke current sQTL pipelines into their basic elements. This led to identifying potential issues/areas we may be able to improv…
biociphers.wordpress.com
November 24, 2025 at 2:32 AM
we published blog posts that give an overview of what's new here: biociphers.wordpress.com/2025/01/20/a... biociphers.wordpress.com/2025/01/28/a...
A nice write up about this work just came out on #alzforum www.alzforum.org/news/researc... and also
www.alzforum.org
November 24, 2025 at 2:32 AM
A nice write up about this work just came out on #alzforum www.alzforum.org/news/researc... and also
Co-localization of sQTLs and AD and PD GWAS variants increased by 64% and 92% respectively compared to current standard, and we experimentally test an sQTL associated with exon 7 in MS4A3 co-localized with AD that disrupts the RBP YBX3 binding site by blocking that region with ASO.
November 24, 2025 at 2:32 AM
Co-localization of sQTLs and AD and PD GWAS variants increased by 64% and 92% respectively compared to current standard, and we experimentally test an sQTL associated with exon 7 in MS4A3 co-localized with AD that disrupts the RBP YBX3 binding site by blocking that region with ASO.
#3: Last but not least, we released a new method for splicing #QTL which uses #MAJIQ under the hood but then adds new statistical modeling. The new MAJIQTL results in major power increase (can be > 2 fold in some cases) with improved control over false positives. www.cell.com/ajhg/fulltex...
A deep dive into statistical modeling of RNA splicing QTLs reveals variants that explain neurodegenerative disease
Current methods for detecting splicing quantitative trait loci (sQTLs) miss many disease-associated
genetic variants. MAJIQTL introduces improved statistical modeling and comprehensive
splicing repres...
www.cell.com
November 24, 2025 at 2:32 AM
#3: Last but not least, we released a new method for splicing #QTL which uses #MAJIQ under the hood but then adds new statistical modeling. The new MAJIQTL results in major power increase (can be > 2 fold in some cases) with improved control over false positives. www.cell.com/ajhg/fulltex...
We identify a new role for DDX55 as APA regulator with an associated mechanism where it can unwind local structure to either hide APA signals or bring those to be in the “right” distance - cool beans! 😎🫘
November 24, 2025 at 2:32 AM
We identify a new role for DDX55 as APA regulator with an associated mechanism where it can unwind local structure to either hide APA signals or bring those to be in the “right” distance - cool beans! 😎🫘
#2: We performed a large scale analysis of alternative polyadenylation ( #APA ) changes after RNA Binding Protein (RBP) Knockdown (KD). The entire APA quantifications across ENCODE is given as a resource you can use but we focus on DDX55. genomebiology.biomedcentral.com/articles/10....
Integrative analysis of RNA binding proteins identifies DDX55 as a novel regulator of 3’UTR isoform diversity - Genome Biology
Background The 3’ untranslated regions (3’UTRs) of mRNAs play a critical role in controlling gene expression and function because they contain binding sites for microRNAs and RNA binding proteins (RBP...
genomebiology.biomedcentral.com
November 24, 2025 at 2:32 AM
#2: We performed a large scale analysis of alternative polyadenylation ( #APA ) changes after RNA Binding Protein (RBP) Knockdown (KD). The entire APA quantifications across ENCODE is given as a resource you can use but we focus on DDX55. genomebiology.biomedcentral.com/articles/10....
Negative selection, effect of mutations...subtypes matter! We validate effect on effect on structure and downstream gene translation of Breast cancer associated variants that either create or disrupt rG4. Transcriptome wide predictions are available for gnomAD as well as online prediction tool
November 24, 2025 at 2:32 AM
Negative selection, effect of mutations...subtypes matter! We validate effect on effect on structure and downstream gene translation of Breast cancer associated variants that either create or disrupt rG4. Transcriptome wide predictions are available for gnomAD as well as online prediction tool
#1: G4mer is an #LLM to predict #RNA G-Quads (rG4) and their subtypes. www.nature.com/articles/s41.... Yes, it's much more accurate... more interestingly, we show current experimental data used by ML/AI methods is biased to short, canonical, rG4 and that not all rG4 are created equal:
G4mer: An RNA language model for transcriptome-wide identification of G-quadruplexes and disease variants from population-scale genetic data - Nature Communications
RNA G-quadruplexes (rG4s) are structures formed in guanine-rich regions of RNA that can serve as crucial regulatory elements in gene expression. Here the authors present an RNA language model for tran...
www.nature.com
November 24, 2025 at 2:32 AM
#1: G4mer is an #LLM to predict #RNA G-Quads (rG4) and their subtypes. www.nature.com/articles/s41.... Yes, it's much more accurate... more interestingly, we show current experimental data used by ML/AI methods is biased to short, canonical, rG4 and that not all rG4 are created equal:
I'll preface, each of these papers represents a different facet of the labs research both in terms of the RNA and the Comp methods, making me very happy and proud about them 😊. So let’s start.. (no particular order)
November 24, 2025 at 2:32 AM
I'll preface, each of these papers represents a different facet of the labs research both in terms of the RNA and the Comp methods, making me very happy and proud about them 😊. So let’s start.. (no particular order)
Forgot to add a link if you want to look it up for more details: rnahorizons.com/rna-horizons...
rnahorizons.com
November 2, 2025 at 12:01 AM
Forgot to add a link if you want to look it up for more details: rnahorizons.com/rna-horizons...
More posts on #RNA_Horizons25 from @grnfluoresceblg.bsky.social, Thank you Michal Lotem, Rotem Karni, Amalia Herszkowicz and all the other organizesrs for the invite and putting such a great event together! 😀🙏 - See you next time!
November 1, 2025 at 11:58 PM
More posts on #RNA_Horizons25 from @grnfluoresceblg.bsky.social, Thank you Michal Lotem, Rotem Karni, Amalia Herszkowicz and all the other organizesrs for the invite and putting such a great event together! 😀🙏 - See you next time!
The above are just from the first ~1.5 day, just headlines, and I still missed a bunch! If I have energy I'll add more later...😅 Bottom line: I learned a *lot* about what's going on in the RNA TX space both academica and industry. High quality talks, in a fabolous venue 😀
November 1, 2025 at 11:58 PM
The above are just from the first ~1.5 day, just headlines, and I still missed a bunch! If I have energy I'll add more later...😅 Bottom line: I learned a *lot* about what's going on in the RNA TX space both academica and industry. High quality talks, in a fabolous venue 😀
12. Riccardo Panella descrived lab and company's approach to target miR-22 affects multiple metabolic pathways. Results compared to GLP-1 show faster, lasting, not affecting lean fat.
November 1, 2025 at 11:58 PM
12. Riccardo Panella descrived lab and company's approach to target miR-22 affects multiple metabolic pathways. Results compared to GLP-1 show faster, lasting, not affecting lean fat.
11. Oxana Beskrovnaya discussed her company's Dyne TX results on dushen and DM1 with targeted delivery to muscle and CNS claims 20fold increase, in mice results last at leats 2mo, in phase 1 much better than existing which is only 1%.
November 1, 2025 at 11:58 PM
11. Oxana Beskrovnaya discussed her company's Dyne TX results on dushen and DM1 with targeted delivery to muscle and CNS claims 20fold increase, in mice results last at leats 2mo, in phase 1 much better than existing which is only 1%.
10. Adrian Krainer talked about SRSF1 regulating Aurka 5' UTR splicing. Normal tissues do not express this isoform, so potentially good target, ASO targeting collapses the SRSF1-AURKA-MYC circuit and induces apoptosis. Maybe combine with KRAS inhibitors, maybe other cancers too.
November 1, 2025 at 11:58 PM
10. Adrian Krainer talked about SRSF1 regulating Aurka 5' UTR splicing. Normal tissues do not express this isoform, so potentially good target, ASO targeting collapses the SRSF1-AURKA-MYC circuit and induces apoptosis. Maybe combine with KRAS inhibitors, maybe other cancers too.
9. Gal Haimovich described Intercellular mRNA transfer via tunneling nanotubes (TNTs) as a potential new way for delivery, about 1% is being transferred want to dev tech to push that further.
November 1, 2025 at 11:58 PM
9. Gal Haimovich described Intercellular mRNA transfer via tunneling nanotubes (TNTs) as a potential new way for delivery, about 1% is being transferred want to dev tech to push that further.