Prime-Editing Screens - Usage Guide

Overview

Decision-grade design and analysis of pooled prime-editor screens. Covers pegRNA design with PRIDICT (Mathis 2023 Nat Biotechnol 41:1151) and PRIDICT2 (Mathis 2025 Nat Biotechnol 43:712, online June 2024); PE chemistry selection (PE2, PE3, PEmax, PE5max); pegRNA architecture (spacer + scaffold + PBS + RTT); chromatin context as a major locus-level determinant; PRIME pooled-screen methodology (Ren 2023); MOSAIC in situ saturation mutagenesis; CRISPResso2 quantification; and cross-modal validation with base-editor screens.

Prerequisites

# PRIDICT2 (web + CLI)
git clone https://github.com/uzh-dqbm-cmi/PRIDICT2
# OR use web interface https://pridict.it/

# CRISPResso2 for amplicon analysis
conda install -c bioconda crispresso2   # not on PyPI

# Helpers
pip install pandas numpy biopython

Required inputs: - Intended variant list with genomic coordinates - Target genome reference (e.g. GRCh38) - Cell line for PE chemistry validation - Chromatin accessibility profile (ATAC-seq) for chromatin-aware filtering

Quick Start

Tell the AI agent what to do: - "Design a pegRNA library for 320 specific ClinVar MLH1, MSH2, MSH6, PMS2 variants. Use PRIDICT2 to score; select top 3 pegRNAs per variant" - "Compare BE vs PE for installing a C->T variant at TP53 R273H. Recommend chemistry based on editing window + bystander" - "Diagnose: my PE screen has 80% scaffold incorporation. Re-design pegRNAs with longer RTT and verify with PRIDICT2" - "Run CRISPResso2 in prime-editor mode on amplicon FASTQ. Report intended-edit %, scaffold incorporation, indel byproduct per pegRNA" - "Cross-validate PE screen hits with parallel BE screen at the same variants; identify PE-only hits in transversion variants"

Example Prompts

pegRNA Library Design

"Build pegRNA library for 500 ClinVar MMR gene variants (MLH1, MSH2, MSH6, PMS2). Score each candidate pegRNA with PRIDICT2. Select 3 best pegRNAs per variant with predicted efficiency >50%. Output peg_library.csv with spacer, PBS, RTT, predicted_efficiency."

"For installing MLH1 c.677A>G as a single intended variant: PE (no bystander) vs BE (no PAM-restriction). Run PRIDICT2 on PE candidates; check editing window for BE; recommend best chemistry."

Screen Analysis

"Run CRISPResso2 in prime-editor mode on my 50-pegRNA pilot. Report intended_edit_pct, scaffold_incorp_pct, indel_pct per pegRNA. Flag pegRNAs with intended_edit <5% for filtering or re-design."

"Apply MAGeCK MLE on my PE screen counts after filtering to efficiency-passing pegRNAs. Aggregate per-pegRNA LFCs to per-variant scores."

Chromatin-Aware Design

"Cross-reference my pegRNA library with K562 ATAC-seq accessibility. Flag pegRNAs targeting closed chromatin (signal <0.5) for empirical pilot validation before library order. Mathis 2025 documented chromatin as a major locus-level determinant beyond PRIDICT sequence prediction, which is why ePRIDICT is meant to be combined with PRIDICT2.0."

MOSAIC / Saturation

"Design MOSAIC saturation library tiling TP53 DNA-binding domain (amino acids 94-294). For each residue, design pegRNAs covering all 19 possible amino acid changes (or at least all functionally relevant). Include variants from ClinVar."

Cross-Validation

"I have parallel BE and PE screens at the same 200 variants. Intersect hits at FDR <0.05; identify high-confidence variants (both methods, same direction). Flag PE-only hits in non-BE-coverable variants (transversions) as genuine PE-unique."

Diagnostics

"PE screen has 30% scaffold incorporation across library. Diagnose: RTT too short relative to PBS, or RT processivity issue with this cell line? Recommend pegRNA re-design or chemistry switch."

"PRIDICT2 predicts 65% efficiency for my top pegRNAs, but pilot CRISPResso2 shows 8%. Identify chromatin context as the cause; cross-reference ATAC-seq."

What the Agent Will Do

Confirm intended variants are PE-installable (NGG PAM within 30 nt; achievable transversion/multi-base if BE-uncoverable)
Decide chemistry: PE2 standard; PEmax for high-efficiency; PE3 with second nick if expression supports
Design pegRNA candidates: spacer + PBS (11-13 nt, 40-55% GC) + RTT (10-20 nt) + scaffold
Run PRIDICT2 to predict per-pegRNA efficiency, indel rate, scaffold incorporation
Filter library to top 3 pegRNAs per variant with predicted efficiency >50%
Cross-reference with ATAC-seq accessibility; flag closed-chromatin loci for pilot
Pilot the library at 20-50 representative loci; verify empirical efficiency matches PRIDICT2
Scale to full screen at MOI 0.3 in PE-expressing cell line
Endpoint amplicon sequencing of each target locus
CRISPResso2 quantification: intended-edit %, scaffold incorp %, indel %
Filter library to >5% intended-edit pegRNAs
Per-variant hit calling with MAGeCK MLE / drugZ
Cross-validate with BE screen if applicable
Output: per-variant fitness, BE-PE concordance, validation strategy

Tips

PRIDICT2 sequence prediction is a starting point, not a guarantee. Chromatin context is a major locus-level determinant beyond sequence (pair PRIDICT2.0 with ePRIDICT); always pilot at representative sites.
For high-stakes variant-function calls, validate with both BE (where applicable) and PE. Concordant hits are high confidence; PE-only at BE-coverable variants are suspect.
PE is the gold standard for transversions (C->G, C->A, G->C, G->A) where BE doesn't work; for these variants, PE is the only choice.
Scaffold incorporation is a design issue, not a chemistry issue. Always check post-PE; if >5%, re-design pegRNAs with longer RTT.
PE3 adds a nicked sgRNA on the opposite strand; this increases editing efficiency but doubles the indel risk. Use PE2 unless PEmax + PRIDICT2 doesn't reach target efficiency.
For genome-scale saturation editing, MOSAIC (2024 preprint, demonstrated on BCR-ABL1 and the IRF1 UTR) is a high-throughput prime-editing option: 1000s of variants per cell line; smaller per-variant cell counts; suitable for drug-resistance variant scanning. For established saturation genome editing of BRCA1, the reference approach remains SGE (Findlay 2018).
Pilot PRIDICT2 predictions in your specific cell line. The deep-learning model was trained on HEK293T (MMR-deficient) and K562 (MMR-proficient); other lines may have different chromatin and RT-activity profiles.
For BE-PE comparison, design parallel libraries targeting the same variants. Concordant hits represent strong biology.

Decision Cheat Sheet

Question	Answer
C->T or A->G at editing window pos 4-8	BE (preferred; higher efficiency)
Multi-base edit	PE
Transversion (C->G, C->A)	PE
Out-of-window single base	PE
Cancer-line variant scanning	PE or BE; cross-validate
iPSC / primary cell variant	PE (no bystander confounding)
Variant for drug-resistance	PE or BE; depends on chemistry
LoF / KO only	Cas9 (no template needed)

Thresholds

Threshold	Value	Rationale
PRIDICT2 prediction threshold	>50% efficiency	Project-chosen cutoff; PRIDICT2 prescribes none
Intended edit % for screen	>5%	Field convention
Scaffold incorporation	<2% (clean); <5% (acceptable)	Empirical
Indel byproduct (PE2)	<3%	Anzalone 2019
PBS GC content	40-55%	PRIDICT2
PBS length	11-13 nt	PRIDICT2
RTT length	10-20 nt	PRIDICT2
Edit position from cut	1-30 nt	Anzalone 2019

Validation Checklist

[ ] Pre-synthesis PRIDICT2 filter (>50%)
[ ] Chromatin accessibility cross-referenced
[ ] Pilot empirical efficiency match >70% of PRIDICT2 prediction
[ ] CRISPResso2 PE mode for quantification
[ ] Scaffold incorporation <5% library-wide
[ ] Per-variant 3+ pegRNAs in library
[ ] BE cross-validation (if BE-coverable variants)

crispr-screens/library-design - pegRNA library design
crispr-screens/base-editing-analysis - Orthogonal BE for variant attribution
crispr-screens/crispresso-editing - CRISPResso2 PE mode
crispr-screens/hit-calling - Per-variant hit calling
crispr-screens/screen-qc - Editing-efficiency QC
variant-calling/variant-annotation - Annotate edited variants
clinical-databases/clinvar-lookup - Pathogenicity annotation