10/ We're continuing to explore generative AI to decode gene regulation through regulatory element modeling and experimental validation, building the next generation of precision gene therapies. If this work excites you or you want to collaborate, send us a message!
9/ Want to read more or try it out?
Paper: nature.com/articles/s41...
Open Access: rdcu.be/eVZ1L
Code: github.com/pinellolab/D...
8/ We're also grateful for the flexible funding to explore this high-risk high-reward research provided by
@rappfoundation.bsky.social and Jason and Keely Krantz through the Krantz Family Center for Cancer Research at
@massgenbrigham.bsky.social.
7/ We're thankful to all of our co-authors who helped bring this work to fruition. We're especially grateful to
@em6wong.bsky.social for the STARR-Seq validation,
@jk-swietek.bsky.social and @bartdeplancke.bsky.social for the EXTRA-Seq experiments, and Wouter Meuleman for his guidance.
6/ We also benchmarked DNA-Diffusion against leading regulatory element design methods. Using classifier-free guidance, our model generates sequences mirroring endogenous elements OR optimized for activation, excelling across multiple cell lines and design objectives.
5/ Of the top 10 DNA-Diffusion sequences, 6 expressed AXIN2 higher than the protective indel. Enformer-predicted CAGE signal at the enhancer correlated strongly with experimental success (Spearman r=0.70), guiding future selection in constrained settings.
4/ Next, we used EXTRA-Seq, a novel experiment system measuring endogenous gene expression, to target AXIN2. An upstream indel drives AXIN2 expression in B-cells, which is associated with improved prognosis in chronic lymphocytic leukemia.
3/ Despite not being trained on expression data, our top DNA-Diffusion sequences consistently outperformed the strongest validated genomic enhancers experimentally.
2/ Previously, we showed DNA-Diffusion generates sequences that recapitulate endogenous grammar and have greater in silico activation potential than training sequences. Now, using STARR-Seq, we've experimentally validated these predictions.
1/ Thrilled to share our DNA-Diffusion paper, co-led by Lucas Ferreira and @ssenan.bsky.social is now out in @NatureGenet! New experimental results: STARR-Seq validation across 3 cell types (episomal) and EXTRA-Seq for endogenous activation of AXIN2, a leukemia protective gene! x.com/lucapinello/...
We are thrilled to announce the first official release (v0.1.8) of #π―π²π±π±π²πΏ, the successor to one of our flagship tool, #π―π²π±ππΌπΌπΉπ! Based on ideas we conceived of long ago (!), this was achieved thanks to the dedication of Brent Pedersen.
1/n
Very proud of our new paper! Great job @mweilert.bsky.social, our experimentalists and modeling collaborator Rosa Martinez-Corral. It was fun to see the story grow and get feedback from various experts. Thank you all!
Applications are open NOW! Review starts December 22, so don't wait.
Apply here: academicjobsonline.org/ajo/jobs/30714
Questions? β ewsc-postdocs@broadinstitute.org
Know someone perfect for this? Please RT and tag them! Let's help spread the word about this opportunity π
What makes this special:
1. YOU choose your mentors (both computational & biomedical)
2. Focus on mathematical foundations, not just applications
2. Ask "why do these algorithms work for biology?" not just "how"
3. Bridge theory βοΈ practice in ways that matter for human health
What they're looking for:
β
Strong background in ML, statistics, or applied math
β
Drive to push biomedical research frontiers
β
Computational/theoretical method development skills
Biology experience? Helpful but NOT required! They value diverse perspectives.
The mission: Create the theoretical & algorithmic foundations to understand the programs of life across scales. You'll choose mentors from faculty across
@broadinstitute.org, @mit.edu and @harvard.edu and collaborate across these amazing institutions.
π¨ Exciting opportunity! The Eric and Wendy Schmidt Center at the Broad Institute of MIT and Harvard is recruiting exceptional postdocs for 2026 at the intersection of ML/math and biomedicine.
Shape the future of computational biology π§΅π
I'm happy to share that our gReLU package is now published in Nature Methods!
www.nature.com/articles/s41...
@zainmunirpatel.bsky.social will also present this afternoon 2.30 to 4.30
Poster 4022W. Systematic benchmarking of computational methods for SVG detection. #spatial #ASHG2025
π
OCT 17 | 2:30-4:30 PM | Poster 5057F
Jayoung Ryu ( @jykr.bsky.social ) presents "SIMBA+: Dissecting genetic variant function through single-cell multiomics integration"
Exciting multiomics integration work! #SingleCell #Multiomics #VariantInterpretation #ASHG2025
π
OCT 16 | 2:30-4:30 PM | Poster 6020T
I'll be at my poster: "CRISPRlungo: Accurate Quantification of Genome Editing Events from Long Reads"
Come chat about long-read sequencing & CRISPR! #GenomeEditing #LongReads #ASHG2025
π
OCT 16 | 10:45-11:00 AM | Room 205ABC
Basheer Becerra presents "Multi-modal functional mapping of non-coding sequences regulating fetal hemoglobin"
Exciting work on therapeutic targets! #FetalHemoglobin #ASHG2025
Zhijian Li ( @zhijianli.bsky.social )presents "Cell2net: A Deep Learning Framework to Decode Genotype-Specific Gene Regulatory Networks from Single-Cell Multi-Omics"
#DeepLearning #GeneRegulation #ASHG2025
π
OCT 15 | 11:00-11:15 AM | Room 258ABC
Logan Blaine presents "PerTurbo: A scalable Bayesian framework for whole-transcriptome differential expression analysis & optimal design of single-cell CRISPR screens"
#CRISPRscreen #SingleCell #ASHG2025
π§¬Excited to be at #ASHG2025! My lab is presenting 5 studies this week. Join us for talks and posters on #SingleCell #CRISPR, #DeepLearning & more! Thread below with all presentation details π
15 years in the making, we confirmed that mitochondria - the powerhouse of the cell - have an unusual localization in patients who experience psychosis (including schizophrenia and bipolar disorders). Youβll never guess what kind of patient cells we used to make this discoveryβ¦ π§΅
I wrote about gene-gene interactions (epistasis) and the implications for heritability, trait definitions, natural selection, and therapeutic interventions. Biology is clearly full of causal interactions, so why don't we see them in the data? A π§΅:
It was indeed fantastic! Thank you so much to all the organizers for supporting this community. π
π The revised version of CORNETO, our unified Python framework for knowledge-driven network inference from omics data, is published in peer reviewed form
π Paper: www.nature.com/articles/s42...
π News & Views: www.nature.com/articles/s42...
π» Code: corneto.org
𧡠Thread π
