4/4: P104:  DeepMirBind: A Deep Learning Approach for MicroRNA Target Site Prediction and Design of Upregulation Oligonucleotides

3/4: P041:  BigRNA: A Large Neural Network Enabling Accurate Oligonucleotide Design and Variant Effect Predictions

2/4: P038:  Predicting novel splicing variants of SYNGAP1 using machine learning predictors: implications for therapeutic oligonucleotides

1/4: P008: Optimized splice-switching oligonucleotide-mediated knockdown of XDH for the treatment of Gout

We're at the #OTS23 Meeting this week -- excited to share updates on our application of #AI to the discovery and development of #oligonucleotide and #RNA therapeutics. Stop by to learn more at the following 4 posters. https://t.co/mCM9aAxpS6

Découvrez ce nouveau travail étonnant du sponsor bronze de Vecteur @deepgenomics et du cofondateur de Vecteur, Brendan Frey. Comme indiqué ci-dessous, l'équipe a développé le premier modèle de base pour les thérapies ARN 👇

Check out this amazing new work from Vector Bronze Sponsor @deepgenomics & Vector co-founder Brendan Frey. As outlined below the team developed the first foundation model for RNA therapeutics 👇

The first foundation #AI model for RNA https://t.co/aNsRMbEuAu Predicting protein and microRNA binding sites, tissue-specific expression, variants -> gene expression and splicing ++ @deepgenomics @frey_brendan and colleagues

16/16 We hope that our model BigRNA, and more RNA foundation models like it, will continue to transform the fields of disease variant interpretation and personalized therapeutics

15/16 Here it’s important to have accurate predictions to lower the number of drugs screened. We analyzed variants in ClinVar that are amenable to this approach.

14/16 Many ultra-rare “N=1” variants could be driven by the sorts of pathogenic changes that we can predict, and amenable to SBO therapy https://t.co/CwMx0Btwuz

13/16 In fact, we consistently found a significant correlation between SBO effects and experimental data for all 18 exons we tried it on

12/16 We predict the activity of the approved drug Spinraza with high accuracy, and we find a strong correlation (r=0.91) for our internal SBO data targeting a pathogenic variant in the ATP7B gene

11/16 We were surprised by the success here, because the model was never trained to predict the effects of these molecules and has never seen data for them

10/16 We went further and asked whether we could use this model to design steric blocking oligos (SBOs) to reverse these disease mutations

9/16 Modeling RNA binding protein sites can also help us discover the mechanisms that are driving these changes

8/16 By explicitly modeling RNA junctions we’ve also created a powerful splicing model that can predict multiple splicing aberrations caused by a single pathogenic variant, such as exon skipping and intron retention

7/16 We complement and extend their work in several ways, such as by explicitly modeling RNA-seq junctions, and the specificity of RNA binding proteins and microRNAs. We apply our work heavily towards discovery of RNA therapeutic candidates

6/16 We acknowledge concurrent work from David Kelley’s team on the Borzoi model and the important technical advance it represents in the modeling of RNA-seq coverage data https://t.co/x4fhNkWKG9

5/16 Following on our recent work on curating disease variants in untranslated regions, we found that modeling changes in RNA-seq offers a powerful way to distinguish these variants from benign ones in general https://t.co/1CtJoyfhhx