OmniGenome 186M EU AI Act Compliance Profile

Bridging Sequence-Structure Alignment in RNA Foundation Models (AAAI 2025)

Model Description

OmniGenome is an advanced RNA foundation model that introduces sequence-structure alignment to genomic modeling. The model bridges the gap between RNA sequences and their secondary structures, enabling bidirectional mappings that improve the flow of genomic information between RNA sequences and structures. With OmniGenome, researchers can achieve improved performance in RNA-related tasks, such as RNA design, secondary structure prediction, and various downstream genomic tasks. It also demon...

• Model type: Transformer-based (52M and 186M parameter versions)
• Languages: RNA sequences and structures
• Pretraining: The model is pretrained on RNA sequences from over 1,000 plant species from the OneKP database. Secondary structures were predicted using ViennaRNA.
• Key Features:
  • Seq2Str (Sequence to Structure) and Str2Seq (Structure to Sequence) mapping
  • RNA design and secondary structure prediction
  • Generalizability to DNA genomic tasks

Intended Use

This model is ideal for:

• RNA secondary structure prediction
• RNA design via structure-to-sequence mapping
• Genomic sequence understanding tasks, such as mRNA degradation rate prediction
• Transfer learning to DNA tasks, including promoter strength prediction, gene expression regression, and more

It is a valuable tool for researchers in RNA genomics, bioinformatics, and molecular biology.

Limitations

OmniGenome is primarily trained on RNA data and its transferability to other genomic data (like human DNA) may require further finetuning. While it demonstrates excellent performance in in-silico experiments, in-vivo validation is yet to be performed.

Training Data

OmniGenome was pretrained on large-scale RNA sequences from the OneKP initiative, which contains transcriptome data from 1,124 plant species. These sequences were processed and cleaned to ensure data quality, and seco

Your obligations as a OmniGenome 186M deployer

Track your progress

Sign up to track obligations, receive deadline reminders, and generate audit packages.

Tags

Languages

Library