GPTomics/bio-crispr-screens-crispresso-editing

Version Compatibility

Reference examples tested with: CRISPResso2 2.2.14+ (pinellolab/CRISPResso2), pandas 2.2+, numpy 1.26+, matplotlib 3.8+.

Before using code patterns, verify installed versions match. If versions differ:

• CLI: CRISPResso --version; CRISPRessoBatch --help; CRISPRessoPooled --help; CRISPRessoWGS --help; CRISPRessoCompare --help
• Python: from CRISPResso2 import ...

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

CRISPResso2 Editing Quantification

"Quantify CRISPR editing from my amplicon sequencing" -> Align amplicon reads against the reference, classify each read as unmodified / NHEJ / HDR / base-edited / prime-edited within the quantification window, and report per-edit-type frequencies, indel size distributions, allele-frequency tables, and substitution-position profiles.

• CLI: CRISPResso -- single amplicon, single sample
• CLI: CRISPRessoBatch -- multi-sample with per-sample parameters
• CLI: CRISPRessoPooled -- multi-amplicon pooled amplicon sequencing
• CLI: CRISPRessoWGS -- off-target quantification from whole-genome BAM
• CLI: CRISPRessoCompare -- pairwise outcome comparison (e.g., treated vs untreated)

Mode Decision Tree

| Experimental design | Mode | Key parameters | |---------------------|------|----------------| | Single amplicon, single sample (e.g. pilot edit validation) | CRISPResso | --amplicon_seq, --guide_seq | | Same amplicon, many samples (e.g. timecourse, dose response) | CRISPRessoBatch | --batch_settings table | | Many amplicons, pooled in one library (e.g. arrayed validation pool) | CRISPRessoPooled | --amplicon_file | | Off-target survey from whole-genome BAM | CRISPRessoWGS | --bam, --reference, --regions_file | | Comparing two CRISPResso runs (e.g. condition A vs B) | CRISPRessoCompare | --crispresso_output_folder_1, _2 | | HDR / knock-in validation | CRISPResso with --expected_hdr_amplicon_seq | Same as base CRISPResso | | Cytosine base editor (C->T) | CRISPResso --base_editor_output | --conversion_nuc_from C --conversion_nuc_to T | | Adenine base editor (A->G) | CRISPResso --base_editor_output | --conversion_nuc_from A --conversion_nuc_to G | | Prime editor (templated edit) | CRISPResso with pegRNA parameters | --prime_editing_pegRNA_spacer_seq, --prime_editing_pegRNA_extension_seq, --prime_editing_pegRNA_scaffold_seq |

Fails when:

• Pooled-amplicon mode applied to amplicons that share primer sequences -- reads get misassigned.
• Base editor mode without specifying --conversion_nuc_from/--conversion_nuc_to -- defaults assume CBE (C->T); ABE runs will misclassify.
• Prime editor mode without --prime_editing_pegRNA_extension_seq -- the RTT template is missing, no edit is detectable.

The Quantification Window

Why this matters for postdoc-level use: CRISPResso classifies reads as "edited" or "unmodified" based on whether modifications fall inside the quantification window (not the whole amplicon). The window is centered on the predicted cut site (Cas9: 3 bp upstream of PAM; Cas12a: 18 bp downstream of PAM) with a default size of 1 (the cut site itself).

# Default Cas9 setup
--quantification_window_size 1                  # 1-bp window at cut site
--quantification_window_center -3               # 3 bp upstream of PAM

# Base editor: widen window to cover editing positions 4-8
--quantification_window_size 10                 # extend to capture position 4-8
--quantification_window_center -10              # center on the editing window

Consequences of mis-sized window:

• Too narrow: misses edits at HDR positions or far bystanders; underestimates editing
• Too wide: includes random sequencing errors; inflates editing rate
• Wrong center: edits at correct position are scored as outside the window

For base editing screens, the convention is --quantification_window_size 10 to cover positions 4-13 from PAM-distal end. For prime editing with multi-base templated edits, widen to encompass the entire edit region.

Single-Amplicon Cas9 Editing

Goal: Quantify indel frequencies and HDR efficiency from a single target site.

Approach: Align FASTQ reads to the reference and (optional) expected-HDR amplicon, classify each read, and report aggregated statistics.

CRISPResso \
    --fastq_r1 sample_R1.fastq.gz \
    --fastq_r2 sample_R2.fastq.gz \
    --amplicon_seq <amplicon_sequence_ref_genome> \
    --guide_seq <20nt_protospacer_no_PAM> \
    --expected_hdr_amplicon_seq <edited_amplicon_for_HDR> \  # OPTIONAL
    --quantification_window_size 1 \
    --quantification_window_center -3 \
    --min_average_read_quality 30 \                          # Phred quality filter
    --output_folder sample_results \
    --name sample_id

# Outputs:
#   sample_results/<name>/CRISPResso_mapping_statistics.txt
#   sample_results/<name>/CRISPResso_quantification_of_editing_frequency.txt
#   sample_results/<name>/Alleles_frequency_table.zip
#   sample_results/<name>/Indel_size_distribution.png
#   sample_results/<name>/Insertion_deletion_substitution.png

Key outputs:

| File | Content | |------|---------| | CRISPResso_mapping_statistics.txt | Reads aligned, reads in quantification window, % alignment, % discarded | | CRISPResso_quantification_of_editing_frequency.txt | % unmodified, % NHEJ, % HDR (if expected), per-edit-class breakdown | | Alleles_frequency_table.zip | Per-allele sequences and frequencies (allele-level resolution) | | Nucleotide_percentage_table.txt | Per-position A/C/G/T/- frequencies (substitutions + deletions) | | Quantification_window_nucleotide_percentage_table.txt | Same, restricted to quantification window (base-editor analysis) | | Reference_modified_chr11_115...png | Sequencing pile-up over amplicon |

Base Editor Quantification

Goal: Distinguish target base conversion from bystander edits and indel byproducts.

Approach: Run CRISPResso with --base_editor_output flag and specify the conversion direction; widen the quantification window to cover the editing window.

# Cytosine Base Editor (CBE): C->T conversion
CRISPResso \
    --fastq_r1 cbe_sample.fastq.gz \
    --amplicon_seq <amplicon_seq> \
    --guide_seq <20nt_protospacer> \
    --base_editor_output \
    --conversion_nuc_from C \
    --conversion_nuc_to T \
    --quantification_window_size 10 \
    --quantification_window_center -10 \
    --output_folder cbe_results \
    --name cbe_sample

# Adenine Base Editor (ABE): A->G conversion
CRISPResso \
    --fastq_r1 abe_sample.fastq.gz \
    --amplicon_seq <amplicon_seq> \
    --guide_seq <20nt_protospacer> \
    --base_editor_output \
    --conversion_nuc_from A \
    --conversion_nuc_to G \
    --quantification_window_size 10 \
    --quantification_window_center -10 \
    --output_folder abe_results \
    --name abe_sample

Reading the output:

| Metric | Where | Interpretation | |--------|-------|----------------| | Target editing % | Quantification_window_nucleotide_percentage_table.txt, target C/A row | Primary endpoint | | Bystander editing % | Same table, other C/A positions in window | Off-target byproduct in window | | Indel rate | CRISPResso_quantification_of_editing_frequency.txt | Cas9-like cut artifacts; should be <5% for clean BE | | Substitution-vs-indel ratio | Derived | Ratio >10 indicates clean BE; <3 indicates cut-mediated mutagenesis instead |

Critical: Bystander editing is intrinsic to base editors (the deaminase acts across a 5-nt window); it is not noise. Report bystander rates alongside target rates. See [[base-editing-analysis]] for variant-call implications.

Prime Editor Quantification

Goal: Quantify pegRNA-templated edits versus indel byproducts and partial edits.

Approach: Provide spacer, extension (PBS + RTT), and scaffold sequences; CRISPResso identifies reads matching the intended edit.

CRISPResso \
    --fastq_r1 pe_sample.fastq.gz \
    --amplicon_seq <amplicon_seq> \
    --guide_seq <20nt_protospacer> \
    --prime_editing_pegRNA_spacer_seq <20nt_protospacer> \
    --prime_editing_pegRNA_extension_seq <RTT+PBS_sequence> \
    --prime_editing_pegRNA_scaffold_seq <scaffold_sequence> \
    --output_folder pe_results \
    --name pe_sample

# Output adds:
#   Prime_editing_outcomes.txt - intended-edit vs scaffold-incorporation vs indel vs unmodified

Reading prime-editor output:

| Metric | Interpretation | |--------|----------------| | Intended edit % | The pegRNA-encoded edit was correctly installed | | Scaffold incorporation % | Reverse transcription read into scaffold instead of stopping at edit; failure mode | | Indel % | Nick-only editing without templated repair; common at low-PE-activity sites | | Unmodified % | Read matches the reference exactly |

A high-quality prime-edit run shows intended-edit fraction >5% and scaffold incorporation <2%. See [[prime-editing-screens]] for pegRNA design rules.

Batch Mode (Multi-Sample, Same Amplicon)

Goal: Process tens to hundreds of samples with same amplicon design (e.g., a timecourse, dose response, or replicate panel).

Approach: Provide a tab-separated batch settings file with per-sample parameters; CRISPRessoBatch runs all in parallel.

# batch_settings.txt (tab-separated, headers required)
# name    fastq_r1                fastq_r2                amplicon_seq    guide_seq
# t0      t0_R1.fq.gz             t0_R2.fq.gz             ACGT...         GUIDE
# t6      t6_R1.fq.gz             t6_R2.fq.gz             ACGT...         GUIDE
# t12     t12_R1.fq.gz            t12_R2.fq.gz            ACGT...         GUIDE
# t24     t24_R1.fq.gz            t24_R2.fq.gz            ACGT...         GUIDE

CRISPRessoBatch \
    --batch_settings batch_settings.txt \
    --batch_output_folder batch_run \
    --skip_failed \
    --n_processes 8

# Outputs:
#   batch_run/CRISPRessoBatch_RUNNING_LOG.txt
#   batch_run/CRISPRessoBatch_quantification_of_editing_frequency.txt  (aggregated)
#   batch_run/CRISPResso_on_<name>/ for each sample

Pooled-Amplicon Mode

Goal: Process multi-amplicon sequencing libraries (e.g., arrayed validation pools).

Approach: Provide an amplicon table with one row per target; CRISPRessoPooled de-multiplexes reads to the correct amplicon.

# amplicons.txt (tab-separated; header may vary by CRISPResso2 version)
# amplicon_name  amplicon_seq    guide_seq
# BRCA1_exon3    ACGT...         GUIDE1
# TP53_exon7     ACGT...         GUIDE2
# KRAS_codon12   ACGT...         GUIDE3

CRISPRessoPooled \
    --fastq_r1 pooled_R1.fastq.gz \
    --fastq_r2 pooled_R2.fastq.gz \
    --amplicon_file amplicons.txt \
    --output_folder pooled_run \
    --n_processes 8

# Outputs:
#   pooled_run/SAMPLES_QUANTIFICATION_SUMMARY.txt
#   pooled_run/CRISPResso_on_<amplicon>/ for each amplicon

Failure mode: Amplicons with shared primer regions get reads assigned to whichever amplicon comes first. Design primers with ≥3-bp distinguishing regions or use unique molecular identifiers.

WGS Off-Target Mode

Goal: Quantify off-target editing from whole-genome sequencing.

Approach: Provide BAM file + reference + BED file of suspected off-target sites; CRISPResso extracts reads from each region and quantifies edits.

CRISPRessoWGS \
    --bam aligned.bam \
    --reference genome.fa \
    --regions_file off_targets.bed \
    --output_folder wgs_run \
    --n_processes 8

Use case: Validate empirically that an in vivo / clinical-grade edit has minimal off-target activity (combine with GUIDE-seq or CIRCLE-seq predicted sites).

Parse Output in Python

Goal: Pull editing metrics into downstream analysis or reports.

Approach: Read the tab-separated quantification files and the JSON metadata.

import pandas as pd
import json
from pathlib import Path

def parse_crispresso(output_dir):
    '''Extract key metrics from CRISPResso output directory.'''
    out = {}
    # Mapping statistics
    map_stats = {}
    with open(Path(output_dir) / 'CRISPResso_mapping_statistics.txt') as f:
        for line in f:
            k, v = line.strip().split('\t')
            map_stats[k] = v
    out['mapping_pct'] = float(map_stats.get('READS_ALIGNED_PERCENTAGE', 'nan'))
    out['reads_aligned'] = int(map_stats.get('READS_ALIGNED', '0'))
    # Editing quantification
    quant = pd.read_csv(Path(output_dir) / 'CRISPResso_quantification_of_editing_frequency.txt', sep='\t')
    out['editing_quant'] = quant.set_index('Amplicon').to_dict()
    # JSON metadata
    info_path = Path(output_dir) / 'CRISPResso2_info.json'
    if info_path.exists():
        out['info'] = json.loads(info_path.read_text())
    return out

Failure Modes

Low alignment rate (<50%)

Trigger: Wrong amplicon sequence (off by one nt, wrong strand, primer-trimmed vs untrimmed). Mechanism: CRISPResso fails to align reads beyond the amplicon edges; discards as unmappable. Symptom: READS_ALIGNED_PERCENTAGE <50%; per-position coverage drops at amplicon edges. Fix: Re-derive amplicon from genome at primer-trimmed boundaries; verify strand orientation; check that primers are NOT included in --amplicon_seq.

High substitution rate but low indel (Cas9 sample)

Trigger: Sample contamination with adjacent amplicon, primer-dimer, or sequencing error inflation. Mechanism: Random substitutions inflate the per-position substitution rate without true indels. Symptom: Substitutions >2% at base positions outside the cut site; alignment metrics look fine. Fix: Increase --min_average_read_quality to 30+; filter contaminating amplicons; check primer-dimer in CRISPResso_RUNNING_LOG.txt.

Bystander C/A editing inflates "editing efficiency"

Trigger: Base-editor sample with wide quantification window; bystander Cs at adjacent positions counted as edits. Mechanism: Default --quantification_window_size 10 includes all positions in editing window; bystander edits are real but distinct from target edit. Symptom: Editing efficiency 80%+ but target SNV is 30%; bystander rate is 50%. Fix: Always read the per-position table (Quantification_window_nucleotide_percentage_table.txt), not just the aggregate. Report target and bystander rates separately. See [[base-editing-analysis]].

Prime editor sample with high scaffold incorporation

Trigger: RTT is too short relative to PBS, or pegRNA stops short. Mechanism: Reverse transcriptase reads past the edit into scaffold sequence; product is detectable but undesired. Symptom: Scaffold incorporation >5%; intended edit efficiency lower than expected. Fix: Re-design pegRNA with longer RTT; verify with PRIDICT2 (see [[prime-editing-screens]]).

MMEJ deletion misclassified as NHEJ

Trigger: Deletions with microhomology at junction; CRISPResso reports them as indels but doesn't distinguish MMEJ. Mechanism: MMEJ creates predictable deletions using flanking microhomologies; biologically distinct from random NHEJ. Symptom: Recurring same-size deletions in allele table (e.g., -7 bp deletion in 30% of reads). Fix: Examine Alleles_frequency_table for over-represented allele patterns; flag MMEJ-mediated deletions for interpretation (these may be inferred from indel hotspots).

Quantitative Thresholds

| Threshold | Value | Source / Rationale | |-----------|-------|--------------------| | Cas9 editing efficiency (functional KO) | >70% indels | Clement 2019 Nat Biotechnol; below this, KO is incomplete | | Indel rate (clean base editor) | <5% | Clement 2019; >5% = unwanted cut activity | | Target conversion (CBE) | >30% | Variable by target; below this, screen power is poor | | Target conversion (ABE) | >30% | ABE typically lower per-base than CBE | | Bystander rate (BE) | <10% acceptable; <5% ideal | Application-dependent; for variant function studies, must be controlled | | Intended-edit % (prime editor) | >5% per-edit | Anzalone 2019 baseline; can be 50%+ at favorable sites | | Scaffold incorporation (PE) | <2% | High-quality pegRNA design | | Alignment rate | >85% | Below this, amplicon design or contamination issue | | Minimum read quality | Phred 30 | Joung 2017 sequencing standard | | Quantification window size (Cas9) | 1 | Clement 2019 default; precise cut-site analysis | | Quantification window size (BE) | 10 | Cover editing window positions 4-13 |

Common Errors

| Error / symptom | Cause | Solution | |-----------------|-------|----------| | Alignment rate <50% | Wrong amplicon sequence | Re-verify; primers should NOT be in amplicon_seq | | All reads "modified" | Misaligned reference | Check amplicon strand; reverse-complement test | | BE shows mostly indels | Cas9 contamination or wrong protein | Re-derive cell line origin; check Cas9 vs nCas9-BE3 | | Inconsistent batch results | Different amplicon_seq per sample | Use CRISPRessoBatch with consistent amplicon | | Pooled-amplicon misassignment | Primer overlap between amplicons | Re-design with ≥3-bp distinguishing regions | | Out-of-window edits ignored | Window too narrow | Increase --quantification_window_size | | Scaffold incorporation high (PE) | RTT too short | Re-design pegRNA | | Allele frequency dominated by 1 read | Low input / clonal | Verify input cell count; rerun if singleton |

References

• Clement K et al. 2019. Nat Biotechnol 37:224. CRISPResso2 algorithm and modes.
• Pinello L et al. 2016. Nat Biotechnol 34:695. Original CRISPResso.
• Anzalone AV et al. 2019. Nature 576:149. Prime editing (PE-1/PE-2/PE-3).
• Komor AC et al. 2016. Nature 533:420. Base editing (BE3).
• Sanson KR et al. 2020. Nat Commun 11:5165. GRACE base-editor screen library.
• Findlay GM et al. 2018. Nature 562:217. Saturation genome editing.

Related Skills

• crispr-screens/base-editing-analysis - Variant-function analysis using CRISPResso2 BE output
• crispr-screens/prime-editing-screens - PRIDICT2 pegRNA design + PE-tiling
• crispr-screens/library-design - sgRNA / pegRNA design for editing screens
• crispr-screens/screen-qc - Editing-efficiency QC for variant interpretation
• variant-calling/variant-annotation - Annotate detected variants downstream
• read-alignment/bwa-alignment - For WGS off-target alignment input