bio-clip-seq-ago-clip-mirna-targets

What this skill does

## Version Compatibility

Reference examples tested with: eCLIP pipeline (Yeo lab), chimeric eCLIP analysis scripts (Manakov 2022), HEAP pipeline (Li 2020), Hyb pipeline (Travis 2014), TargetScanHuman 8.0, miRDB 6.0, samtools 1.19+, bedtools 2.31+, pyHyb 0.4+.

Before using code patterns, verify installed versions match. If versions differ:
- Python: `pip show <package>` then `help(module.function)` to check signatures
- CLI: `<tool> --version` then `<tool> --help` to confirm flags

If code throws unexpected errors, introspect the installed package and adapt the example to match the actual API rather than retrying.

# AGO-CLIP and miRNA Target Identification

**"Identify direct miRNA-target interactions experimentally"** -> Use Argonaute (AGO1-4) CLIP-seq variants to map miRNA-binding sites on mRNAs, then resolve which miRNA pairs with each site. Three approaches: (a) standard AGO-CLIP recovers AGO-bound sites but cannot say which miRNA; (b) chimeric methods (CLEAR-CLIP, chimeric eCLIP / miR-eCLIP) ligate the miRNA to its target during library prep, producing miRNA-mRNA chimeric reads that unambiguously assign miRNA-target pairs; (c) HEAP uses HaloTag-Ago2 for in vivo profiling. The chimeric methods are the gold standard for direct miRNA-target identification; standard AGO-CLIP must be combined with computational seed-matching (TargetScan, miRDB) to infer miRNA pairing. Resolution: chimeric reads pinpoint single miRNA-target pairs; AGO-only CLIP identifies "AGO-binding sites" of which a subset are miRNA targets.

- CLI (chimeric eCLIP / miR-eCLIP processing): custom pipeline starting from eCLIP-style preprocessing + chimeric-read identification + miRNA-mRNA junction extraction
- CLI (CLEAR-CLIP custom Moore 2015 pipeline): Hyb (Travis 2014) for chimera analysis
- CLI (Hyb pipeline): `hyb run_hyb peaks.bam mature_miRNA.fa human.tab.gz` to find miRNA-mRNA chimeras
- CLI (HEAP analysis): standard HITS-CLIP processing pipeline + Halo-Ago2 capture details
- Python (seed-pairing analysis on AGO CLIP peaks): scan peaks for canonical 7mer-m8, 7mer-1A, 8mer, 6mer seeds + 3' UTR position + miRNA expression filter

The Yeo lab miR-eCLIP / chimeric eCLIP (Manakov 2022) is the modern depth-improved version of chimeric AGO-CLIP, enriching for chimeras of specific miRNAs of interest (30-175x enrichment via PCR or on-bead probe capture). For comprehensive miRNA-target mapping, miR-eCLIP combined with eCLIP-seq-style normalization is the state-of-the-art.

## Methods Taxonomy

| Method | Year | miRNA-target pairing | Chimera enrichment | Strength | Fails when |
|--------|------|---------------------|--------------------|----------|------------|
| HITS-CLIP for AGO | 2009 (Chi) | Indirect (computational seed) | None | Original; widely cited | Cannot assign miRNA without computational prediction |
| PAR-CLIP for AGO | 2010 (Hafner) | Indirect | None | T->C signature at CL position | Restricted to 4SU-permissive cells |
| AGO-CLEAR-CLIP (Moore 2015) | 2015 | Direct (chimera) | None (incidental) | First direct miRNA-target chimera method | Chimeric reads only 1-5% of library; deep sequencing needed |
| CLASH (Helwak 2013) | 2013 | Direct (chimera) | None | First general chimera method; pan-Argonaute | Lower chimera rate than CLEAR-CLIP |
| HEAP (Li 2020) | 2020 | Indirect (with chimeric step) | None | HaloTag-Ago2 in vivo mouse strain | Mouse only; requires transgenic model |
| chimeric eCLIP / miR-eCLIP (Manakov 2022) | 2022 | Direct (chimera) | 30-175x enriched | Deepest miRNA-target chimera profiling | Specialized library prep |
| AGO HITS-CLIP + targeted chimeric (Bracken et al; verify exact venue/year) | -- | Direct | Yes | Per-miRNA enrichment via probe capture | Older; superseded by miR-eCLIP |
| AGO-IP-RNA-seq (Karginov 2007) | 2007 | Indirect | None | Earliest; predecessor of CLIP for AGO | No crosslinking; misses transient targets |

Methodology evolves; verify the Manakov 2022 / Bracken 2024 papers for current chimeric eCLIP best practice. As of 2024, miR-eCLIP is the canonical approach for deep miRNA-target profiling.

## Critical Choice: Chimeric vs Computational miRNA-Target Pairing

Two fundamentally different strategies:

**Chimeric methods (CLEAR-CLIP, chimeric eCLIP / miR-eCLIP, CLASH):** During library prep, a ligation step covalently joins the miRNA to its target mRNA, producing chimeric reads (miRNA at 5' + target mRNA at 3'). The miRNA-target pair is read directly from the sequence. Pro: direct evidence of binding interaction; no inference. Con: chimera rate is 1-5% of library by default (enriched to 30-175x with miR-eCLIP for specific miRNAs); deep sequencing or enrichment needed.

**Computational pairing (HITS-CLIP / PAR-CLIP + seed-matching):** Standard AGO CLIP identifies AGO-bound peaks; downstream computational scanning matches each peak against canonical miRNA seeds (7mer-m8, 7mer-1A, 8mer) from TargetScan, miRDB, or DIANA databases. Pro: any AGO CLIP data can be analyzed; no special library prep. Con: indirect; assigns miRNAs based on canonical seed rules, missing non-canonical interactions (3' compensatory, central pairing).

The Bartel lab CLEAR-CLIP analysis revealed substantial 3' auxiliary pairing beyond canonical seeds: ~50% of miRNA-target interactions have weak or non-canonical seed matching but strong 3' supplementary pairing. Chimeric methods recover these; computational seed-only inference misses them.

| Goal | Method |
|------|--------|
| Direct miRNA-target pair identification | Chimeric eCLIP / miR-eCLIP |
| Specific miRNA's targets (deep) | miR-eCLIP with probe-capture enrichment |
| All AGO-binding sites (any miRNA) | Standard AGO eCLIP / HITS-CLIP |
| In vivo mouse tissue | HEAP (Halo-Ago2 mouse) |
| Pan-Argonaute interactome | CLASH or chimeric eCLIP |
| Initial discovery / cost-conscious | AGO HITS-CLIP + TargetScan |
| Non-canonical / 3'-compensatory miRNA pairing | Chimeric methods (CLEAR-CLIP) |
| Comparison across species | TargetScan + AGO HITS-CLIP (computational) |
| Validate specific miRNA-target prediction | miR-eCLIP with that miRNA's probe |

## miRNA-Target Pairing Rules

Computational seed-matching against TargetScan / miRDB requires understanding the canonical miRNA-target pairing rules:

| Seed type | Pairing positions (miRNA nt) | Position 1 | Pro / Con |
|-----------|-------------------------------|------------|-----------|
| 8mer | 2-7 + position 8 + A at position 1 | A required | Strongest; most conserved targets |
| 7mer-m8 | 2-7 + position 8 (no A1 requirement) | Any | Strong; common |
| 7mer-A1 | 2-7 (no position 8) + A at position 1 | A required | Moderate; common |
| 6mer | 2-7 | Any | Weak; very common (many false positives) |
| 6mer-A1 | 2-6 + A at position 1 | A required | Weak |
| 3'-compensatory | Weak 6mer + strong 3' UTR pairing 12-17 | Any | Discovered via CLEAR-CLIP; misses in seed-only methods |
| Central pairing | Positions 4-15 with no seed | Any | Rare; cleavage rather than translational repression |

For TargetScan integration: download the TargetScanHuman 8.0 conserved-site predictions; filter for 7mer-8mer (drop 6mer if too noisy); cross-reference with the CLIP peak BED of the analysis.

## Chimeric eCLIP / miR-eCLIP Workflow

**Goal:** Recover miRNA-mRNA chimeras from AGO chimeric eCLIP / miR-eCLIP libraries and produce a per-miRNA target list suitable for direct biological interpretation.

**Approach:** Apply eCLIP-style preprocessing, then run Hyb in chimera (`type=mim`) mode with bowtie2 alignment (required for short 21-23 nt miRNA sequences), filter chimeras to human mRNA targets, intersect with miRNA-expression atlas (filter > 100 TPM in matched cell type), and validate top targets against TargetScan conserved 7mer-m8 / 8mer predictions.

```bash
# Step 1: eCLIP-style preprocessing (see clip-seq/clip-preprocessing)
umi_tools extract --bc-pattern=NNNNNNNNNN \
    --stdin=R1.fq.gz --read2-in=R2.fq.gz \
    --stdout=R1.umi.fq.gz --read2-out=R2.umi.fq.gz

cutadapt -a AGATCGGAAG