m6A CLIP - Usage Guide

Overview

Map N6-methyladenosine at single-nucleotide resolution. Five method families: antibody-based UV-CL (miCLIP/miCLIP2), antibody-free chemical (GLORI 2023, the new gold standard for stoichiometric m6A), enzyme-fusion editing (DART-seq APOBEC1-YTH), nanopore direct RNA (m6Anet, nanocompore, EpiNano), and peak-level MeRIP-seq. DRACH motif constrains 70-90% of sites but is not exclusive. Cross-method discordance is real: DART 44% in DRACH; GLORI vs miCLIP2 ~60% concordance. Triangulate across orthogonal methods for high-confidence sites.

Prerequisites

conda install -c bioconda samtools bedtools
pip install m6anet nanopolish-eventalign
# miCLIP2 m6Aboost: https://github.com/ZarnackGroup/m6Aboost
# GLORI: https://github.com/liucongcas/GLORI-tools
# Bullseye DART: https://github.com/mekoulnik/Bullseye

Quick Start

Tell your AI agent: - "Map m6A with miCLIP2 + m6Aboost ML on Mettl3-KO calibrated data" - "Run GLORI for stoichiometric m6A fraction per site" - "DART-seq with Bullseye, subtracting APOBEC1-only control" - "m6Anet on nanopore direct RNA at >= 20 coverage" - "MeRIP-seq peaks with exomePeak2" - "Triangulate miCLIP2, GLORI, m6Anet for high-confidence sites" - "Why is DART editing in non-DRACH? Off-target APOBEC1"

Example Prompts

miCLIP2

"Run miCLIP2 pipeline: eCLIP-style preprocessing + PureCLIP single-nt + m6Aboost ML"

"Filter at m6Aboost score >= 0.5; expect 10-50k high-confidence sites"

GLORI

"GLORI-tools on pre/post conversion BAMs; per-A stoichiometric m6A fraction"

"Filter at coverage >= 20 and m6A_fraction >= 0.1"

DART-seq

"Bullseye for DART editing; subtract APOBEC1-only control; filter for DRACH"

m6Anet (Nanopore)

"nanopolish eventalign on aligned BAM, then m6anet dataprep + inference"

Reconciliation

"Cross-reference miCLIP2 sites with GLORI - if both call, high confidence"

"miCLIP2 calls 100k sites - far too many; apply m6Aboost"

"DART 44% in DRACH - subtract APOBEC1 control, filter for DRACH"

What the Agent Will Do

Pick method based on requirements: GLORI (stoichiometric), miCLIP2 (eCLIP-compatible), m6Anet (nanopore), MeRIP (cheap initial)
For miCLIP2: eCLIP preprocessing -> alignment -> single-nt CL -> m6Aboost ML
For GLORI: pre/post conversion BAMs -> GLORI-tools -> per-A stoichiometric fraction
For DART-seq: Bullseye with APOBEC1-only control; filter DRACH
For m6Anet: nanopolish eventalign -> m6anet inference
Triangulate across 2+ orthogonal methods
Report DRACH-context fraction separately from non-DRACH
Document method-specific limitations

Tips

GLORI is the new (2023) gold standard. Stoichiometric, antibody-free.
miCLIP2 + m6Aboost is the eCLIP-compatible method. Trained on Mettl3 KO.
m6Anet for nanopore / isoform context. High AUC on HEK293T; needs >= 20 cov per DRACH.
DART-seq needs APOBEC1-only control. 30-50% off-target without it.
DRACH covers 70-90% of m6A. Filter selectively; some real sites are non-DRACH.
Method discordance is real. 60-75% concordance across methods.
MeRIP-seq is peak-level. Cannot pinpoint single A.
Mettl3 KO is the negative control. Use it for any antibody-based method.
GLORI degrades long RNAs. Use shorter conversion or other method.
Triangulate 2+ methods for publication. Single method = lower confidence.

clip-seq/clip-preprocessing - miCLIP2 preprocessing
clip-seq/clip-alignment - miCLIP2 alignment
clip-seq/crosslink-site-detection - miCLIP2 CL detection
clip-seq/clip-peak-calling - MeRIP-seq exomePeak2
clip-seq/stamp-antibody-free - STAMP / DART antibody-free
long-read-sequencing/nanopore-methylation - Native nanopore m6A
long-read-sequencing/basecalling - dRNA basecalling
epitranscriptomics/m6a-peak-calling - MeRIP peaks
epitranscriptomics/m6a-differential - Differential m6A
epitranscriptomics/m6anet-analysis - m6Anet workflow