Appearance
Borzoi Mini K562 RNA-seq
Borzoi predicts RNA-seq-like signal from long DNA sequence windows. Liatir starts with the official Mini Borzoi K562 RNA-seq model because it is a smaller managed entry point into the Borzoi family.
What it does
The model reads a long genomic sequence window and predicts a K562 RNA-seq signal. With an optional VCF file, Liatir compares reference and alternate windows to estimate signal changes for selected variants.
When to use it
Use this model when you want to explore RNA-seq signal prediction from DNA sequence with a managed local Borzoi runtime.
Inputs in Liatir
- Reference FASTA/FA/FNA file, or an inline DNA sequence.
- Optional VCF or VCF.GZ file.
- Output head:
Human. - Target index.
- Maximum variants to score.
Outputs
Liatir writes:
- predicted signal CSV;
- BED signal track;
- optional variant score CSV and BED track;
- JSON summary;
- Results panels and provenance.
Hardware and installation
Borzoi is TensorFlow-based and uses very long input windows. Liatir installs it as its own AI Model box and downloads the official Mini Borzoi K562 RNA-seq fold 0 weights, parameters, and targets.
The official Borzoi documentation recommends Python 3.10 with TensorFlow 2.15.x. CPU can run small checks, but GPU is strongly preferred.
Limits and cautions
This is a focused Mini Borzoi model, not the full multi-replicate Borzoi stack. Use it to validate workflows and explore K562 RNA-seq signal prediction before moving to larger Borzoi runs.
Because this model is K562-focused, interpret outputs as a model-specific signal example rather than a universal regulatory prediction for every tissue or cell type.