VAdediwura/aptamer-designer-v7

AptamerDesigner V7

Design DNA aptamers against any protein target using the V7 transformer model.

What it does

Given a protein target, the pipeline:

1. 1.Embeds the protein sequence with ESM2-35M (Facebook Research)
2. 2.Runs a genetic-algorithm-based aptamer design search
3. 3.Scores all candidate aptamers with the trained AptamerScorerV7 model
4. 4.Returns the top aptamers ranked by composite binding score

Inputs

Preset Targets (APOBEC3)

Pre-configured for APOBEC3A, APOBEC3B, and APOBEC3C. Just select a target and click Design Aptamers.

Custom Target

• —FASTA sequence of your target protein (single sequence)
• —Interface residue positions — comma-separated, 0-indexed from the sequence start. These are the residues defining the binding pocket (e.g. active-site residues, catalytic loops, zinc-coordination residues from the PDB structure or literature).

Output columns

Model

• —Architecture: AptamerScorerV7 — transformer-based protein–DNA cross-attention model
• —Parameters: 28,144,129
• —Training data: 70,876 protein-DNA binding pairs from PDB
• —Protein encoder: ESM2-35M interface embeddings (480-dim)
• —DNA encoder: Token embedding + transformer encoder

Citation

If you use this tool, please cite the AptamerDesigner V7 pipeline and the following:

• —ESM2: Lin et al., Science 2023
• —ViennaRNA: Lorenz et al., Algorithms for Molecular Biology 2011

Notes

• —Design runs take 2–5 minutes depending on the number of candidates explored.
• —For APOBEC3 targets, pre-computed protein embeddings are used for faster startup.