About this tool

This tool predicts miRNA binding affinity to RNA targets and estimates competitive effects of optional competitor molecules. It fuses sequence features with optional 3D structure-derived features and evaluates inputs in RAM-safe batches.

How it works

• Input your miRNA(s) as multi-FASTA. You may also provide one or more targets (multi-FASTA) and zero or more competitors (multi-FASTA).
• For each competitor, the backend evaluates all targets against all miRNAs (baseline + with-competitor), so every triple (miRNA, target, competitor) is covered.
• For every molecule (miRNA, target, competitor), the backend applies a unified sequence cascade: FASTA → PDB nucleic acid → AA→NT back-translation so analytics always run on nucleotides.
• Advanced options include auto-trimming long miRNAs to a mature-like window and lossy AA→NT back-translation for protein or protein-only PDB inputs (when enabled).
• The backend validates inputs, encodes sequences and any usable structural features, and performs model inference to produce affinity scores.

1. FASTA wins when present. If a FASTA record exists for a molecule, that nucleotide sequence is trusted for scoring. Any matching PDB/mmCIF is paired as its 3D partner (for viz and optional structure features).
2. PDB-only is allowed. If no FASTA exists for a molecule but a PDB/mmCIF or PDB ID is provided, the backend will derive a nucleotide sequence: nucleic chains are used directly; protein-only chains use AA→NT back-translation when enabled.
3. PDB never silently disables scoring. If a PDB cannot provide a usable sequence (e.g., unknown polymer and no FASTA), it is kept for 3D viewing and clearly marked as “not used for scoring” instead of causing a hard failure for other valid inputs.
4. FASTAn + PDB together behave like one molecule. When IDs match, FASTA supplies the NT sequence, PDB supplies the structure; when they do not, FASTA-based scoring still runs and the PDB is treated as an independent visualization entry.

• miRNA: multi-FASTA required (each sequence needs a header like >hsa-let-7a-5p). Full header text is preserved.
• Targets & Competitors: multi-FASTA supported. Paste RNA/DNA or protein (AA). Protein is back-translated if enabled.
• Range-aware IDs: Use Header:Start-End (e.g., TP53_3UTR:90-150). Seed Sites show local and global coordinates.
• 3D (PDB/mmCIF): upload files or provide PDB IDs. Filenames/IDs are matched to FASTA headers with tolerant rules. Chain hint supported: >TP53_3UTR|chain=A.

Results interpretation

• Baseline affinity: predicted binding strength without competitor.
• With competitor: predicted binding strength considering the competitor.
• Competitive effect: Baseline − With-competitor (higher implies stronger competitive interference).