← Về thư mục
📄 / / profiles / r-and-d / raw_evidence / OpenAI4S / skills / bioskills / bio-clip-seq-binding-site-annotation / references / usage-guide.md

Binding Site Annotation - Usage Guide

Overview

Annotate CLIP-seq peaks to RNA features (5' UTR, CDS, 3' UTR, intron, splice site, ncRNA, repeat element). Interpretation is RBP-class-specific: splicing factors at splice junctions, HuR/PUM2 in 3' UTRs, EIF3 in 5' UTRs, FASTKD2 on chrM, repeat-binding RBPs (MATR3, HNRNPK) on Alu/LINE/LTR. ChIPseeker is the fast tool; RBP-Maps (Yeo lab) is mandatory for splicing factors; RepeatMasker overlap is a separate axis. Default ChIPseeker tssRegion=c(-3000,3000) over-extends for CLIP - tighten to c(-100,100).

Prerequisites

BiocManager::install(c('ChIPseeker','RCAS','GenomicFeatures',
                       'TxDb.Hsapiens.UCSC.hg38.knownGene'))

conda install -c bioconda bedtools deeptools rseqc
# RBP-Maps (Yeo)
git clone https://github.com/YeoLab/rbp-maps

Quick Start

Tell your AI agent what you want to do: - "Annotate peaks to 5' UTR / CDS / 3' UTR / intron with ChIPseeker, transcript-level, tight TSS region" - "Build the RBP-Maps splicing regulatory map for my PTBP1 eCLIP" - "Add a repeat-element axis to my annotation (Alu / LINE / LTR / SINE)" - "Metagene of CLIP signal across 3' UTRs with deepTools computeMatrix" - "Why do 50% of my peaks fall in 'Promoter (2-3kb)'? Tighten TSS region" - "Sum of region counts > total peaks - enforce hierarchy" - "FASTKD2 should bind chrM - my TxDb excludes it"

Example Prompts

Global Region Distribution

"ChIPseeker annotation with tssRegion=c(-100,100), level=transcript, then plotAnnoPie and plotDistToTSS"

"What fraction of my HuR peaks are in 3' UTRs versus introns? Expected > 50% 3' UTR"

Splicing Factor Maps

"Run RBP-Maps for PTBP1 - generate the 1400 nt cassette-exon regulatory metagene"

"Need cassette exon BED from ENCODE shRNA-KD RNA-seq before RBP-Maps will produce signal"

Repeat-Element Axis

"MATR3 binds LINE-1; intersect peaks with RepeatMasker BED and report repeat-class fraction"

"My peaks are 30% Alu - is this real or alignment artifact? Compare to ENCODE Alu binders"

Metagene

"deepTools metagene of CLIP signal scaled across 3' UTRs - look for stop-codon-proximal peak"

"RSeQC geneBody_coverage.py to check for 5' vs 3' positional bias"

Diagnostics

"Many peaks labeled 'Distal Intergenic' - is my TxDb missing transcripts?"

"RBP-Maps metagene is flat - I don't have the cassette exon table"

What the Agent Will Do

  1. Choose ChIPseeker (fast global) + RCAS (RNA-aware details) + RBP-Maps (splicing) based on RBP class
  2. Set tssRegion=c(-100,100), level='transcript', customized priority hierarchy
  3. Annotate peaks; report per-region pie, distance-to-TSS, distance-to-stop-codon
  4. Cross-check expected vs observed dominant region (HuR -> 3' UTR; PTBP1 -> intron; FASTKD2 -> chrM)
  5. Add repeat-element axis via RepeatMasker intersect (-s for strand)
  6. For splicing factors: generate the RBP-Maps cassette-exon regulatory metagene
  7. Flag anomalies: too many "Promoter", too many "Intergenic", missing chrM, repeat over-counting

Tips