[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7ECFN38CH5PC3Y7JEAHHGJC":3,"skill-same:01M7ECFN38CH5PC3Y7JEAHHGJC":62},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":14,"repositoryHtmlUrl":33,"repositoryHomepageUrl":33,"ownerHtmlUrl":34,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":35,"install":36,"resources":49},"01M7ECFN38CH5PC3Y7JEAHHGJC","bio-crispr-screens-base-editing-analysis",null,"Analyzes base-editing screens for variant function. Covers library design (Hanna 2021 ClinVar-scale CBE screen benchmarked on BRCA1\u002F2, Cuella-Martin 2021 DDR saturation), CBE vs ABE chemistry choice (BE3\u002FBE4 vs ABE7.10\u002FABE8.20\u002FABE8e), editing-window math (positions 4-8 from PAM-distal end; 4-7 for ABE7.10), bystander-edit quantification and the variant-call ambiguity it creates, sgRNA-efficiency filtering before hit calling, indel byproduct interpretation, the substitution-vs-indel diagnostic, variant annotation against ClinVar \u002F COSMIC, and the Broad be-validation-pipeline. Use when designing a BE variant screen, choosing CBE vs ABE for a specific edit, interpreting bystander-confounded hits, distinguishing functional signal from indel artifact, integrating CRISPResso2 output with screen scoring, or deciding BE vs PE for SNV installation.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-crispr-screens-base-editing-analysis\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-crispr-screens-base-editing-analysis\ndescription: Analyzes base-editing screens for variant function. Covers library design (Hanna 2021 ClinVar-scale CBE screen benchmarked on BRCA1\u002F2, Cuella-Martin 2021 DDR saturation), CBE vs ABE chemistry choice (BE3\u002FBE4 vs ABE7.10\u002FABE8.20\u002FABE8e), editing-window math (positions 4-8 from PAM-distal end; 4-7 for ABE7.10), bystander-edit quantification and the variant-call ambiguity it creates, sgRNA-efficiency filtering before hit calling, indel byproduct interpretation, the substitution-vs-indel diagnostic, variant annotation against ClinVar \u002F COSMIC, and the Broad be-validation-pipeline. Use when designing a BE variant screen, choosing CBE vs ABE for a specific edit, interpreting bystander-confounded hits, distinguishing functional signal from indel artifact, integrating CRISPResso2 output with screen scoring, or deciding BE vs PE for SNV installation.\ntool_type: mixed\nprimary_tool: CRISPResso2\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"crispr-screens\u002Fbase-editing-analysis\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: CRISPResso2 2.2.14+, BE-Hive 1.0+ (BE prediction), pandas 2.2+, biopython 1.83+, numpy 1.26+, scipy 1.12+, scikit-learn 1.4+; Broad be-validation-pipeline notebooks (repo HEAD).\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `CRISPResso --version`\n- Python: `pip show CRISPResso2`; BE-Hive is a GitHub clone (maxwshen\u002Fbe_predict_bystander), not a PyPI package\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.\n\n## Base Editing Screen Analysis\n\n**\"Analyze my base-editor variant-function screen\"** -> Quantify per-sgRNA target-base conversion, bystander rate, and indel byproducts from amplicon sequencing; filter on editing efficiency; map each sgRNA to its intended SNV (target + bystander pattern); compute per-variant fitness from the screen log-fold change; reconcile target vs bystander variant attribution; annotate against ClinVar \u002F COSMIC.\n\n- CLI: `CRISPResso --base_editor_output` for per-amplicon BE quantification\n- CLI: Broad `be-validation-pipeline` for end-to-end pooled-screen analysis with editing-efficiency filtering\n- Python: `BE-Hive` (Arbab 2020) for editing-efficiency prediction; clone maxwshen\u002Fbe_predict_bystander and import via sys.path\n- Web: `BE-Designer` (Hwang 2018, RGEN Tools) for variant-encoding sgRNA design\n\n## Base Editor Chemistry Selection\n\n| Editor | Reaction | Editing window | Indel byproduct rate | When to use |\n|--------|----------|----------------|----------------------|-------------|\n| BE3 (Komor 2016) | C->T (also G->A on opposite strand) | Pos 4-8 from PAM-distal end | 5-10% | Original; superseded |\n| BE4 \u002F BE4max (Koblan 2018) | C->T | Pos 4-8 | \u003C5% | CBE standard |\n| eA3A-BE3 | C->T narrow specificity | Pos 5-7 | \u003C5% | Specifically TC contexts (eA3A prefers TC) |\n| ABE7.10 (Gaudelli 2017) | A->G (T->C opposite strand) | Pos 4-7 | \u003C2% | First ABE; slow at non-TA contexts |\n| ABE8.20 (Gaudelli 2020) | A->G | Pos 4-8 | \u003C2% | Modern ABE; high activity |\n| ABE8e (Richter 2020) | A->G | Pos 4-8 | \u003C2% | Highest editing activity; more processive than ABE7.10 |\n| evoCDA-BE | C->T (broader) | Pos 1-9 | 5-10% | Larger editing window; more bystander |\n| CGBE1 (Kurt 2021) | C->G | Pos 5-7 | 5-10% | C-to-G transversion; rare use |\n| GBE (Zhao 2021) | C->G or C->A | Pos 4-7 | 5-10% | Transversions; less mature |\n\n**Decision rule:** For a target SNV at position 4-8 of a candidate spacer with no bystander Cs\u002FAs in the same window, BE3-BE4 or ABE7.10 is sufficient. For high-throughput variant scanning where bystander tolerance must be minimized, use eA3A-BE3 (TC contexts only) for C->T, or ABE7.10 rather than ABE8e\u002FABE8.20 for A->G -- its 4-7 window is the narrowest ABE.\n\n## Editing Window Math\n\n**Why this matters for postdoc-level use:** Base editors are tethered to dCas9 (or nCas9) and the deaminase acts on the displaced ssDNA \"R-loop\" formed when Cas9 binds. The deaminase has a fixed reach -- positions 4-8 from the PAM-distal end of the protospacer for canonical BE3\u002FBE4, and 4-7 for ABE7.10. Outside this window, editing efficiency drops by 10-50x.\n\n```\nPAM-distal end                                                            PAM-proximal\n   |                                                                          |\n   1  2  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 20    NGG\n                  ^^^^^^^^^^^\n                  Canonical editing window (positions 4-8)\n\n   For BE4max: positions 4-8 are 5-50x more efficient than positions 1-3 or 9-13 (ABE7.10: 4-7)\n   For SpABE8e: positions 4-8 (Richter 2020), matching the corresponding CBEs rather than ABE7.10's narrower 4-7\n   For evoCDA-BE: window 1-9 (broader; more bystander)\n```\n\n**Critical implication for variant interpretation:** If the intended edit is at position 5 and there is an additional editable C\u002FA at position 7, both will be edited in the same molecule. The screen scores the *combination* of edits, not the intended one alone. This is bystander confounding.\n\n## sgRNA Library Design for BE Screens\n\n**Goal:** Tile editing-window-positioned spacers across a protein region of interest to enable variant scanning.\n\n**Approach:** For each amino acid in the target region, find NGG-adjacent spacers where the SNV-of-interest base falls in editing positions 4-8 with minimal bystander C\u002FA in the same window. Annotate each spacer with the predicted amino acid changes (target + bystander).\n\n```python\nimport pandas as pd\nimport re\nfrom Bio.Seq import Seq\n\ndef find_be_spacers(cds_sequence, cds_protein_start, target_aa, target_base='C', editor='BE4max'):\n    '''Find sgRNAs that place target_base in editor-specific window at target_aa.\n    Returns spacers with bystander annotation.\n\n    Args:\n        cds_sequence: nucleotide CDS (translated frame 1)\n        cds_protein_start: amino acid number of CDS start (usually 1)\n        target_aa: amino acid number to install variant (e.g., 130 for residue 130)\n        target_base: 'C' (CBE) or 'A' (ABE)\n        editor: 'BE3', 'BE4max', 'eA3A-BE3', 'ABE7.10', 'ABE8.20', 'ABE8e', 'evoCDA-BE'\n\n    Returns: DataFrame with spacer, position-in-cds, target-base-position-in-spacer,\n             bystander_positions, predicted_aa_changes\n    '''\n    # Editor-specific editing window (positions from PAM-distal end of spacer)\n    window_by_editor = {\n        'BE3': (4, 8),       'BE4max': (4, 8),    'eA3A-BE3': (5, 7),\n        'ABE7.10': (4, 7),   'ABE8.20': (4, 8),   'ABE8e': (4, 8),     # SpABE8e matches CBE window (Richter 2020)\n        'evoCDA-BE': (1, 9),\n    }\n    window_lo, window_hi = window_by_editor[editor]\n    aa_index = target_aa - cds_protein_start  # 0-indexed in protein\n    aa_start_nt = aa_index * 3                # nt offset in cds\n    candidates = []\n    spacer_len = 20\n    pam_pattern = re.compile(r'(?=([ACGT]GG))')\n    for strand, seq in [('+', cds_sequence), ('-', str(Seq(cds_sequence).reverse_complement()))]:\n        for pam_match in pam_pattern.finditer(seq):\n            pam_pos = pam_match.start()\n            spacer_start = pam_pos - spacer_len\n            if spacer_start \u003C 0:\n                continue\n            spacer = seq[spacer_start:pam_pos]\n            # Editor-specific window from PAM-distal end (1-indexed)\n            # Find all editable bases in window\n            edit_bases_in_window = []\n            for i, b in enumerate(spacer[window_lo-1:window_hi], start=window_lo):\n                if b == target_base:\n                    edit_bases_in_window.append(i)\n            if not edit_bases_in_window:\n                continue\n            # Annotate which edits hit the target_aa codon\n            target_codon_start = aa_start_nt\n            target_codon_end = target_codon_start + 3\n            target_position_in_spacer = []\n            for i in edit_bases_in_window:\n                genomic_pos = spacer_start + i - 1\n                if target_codon_start \u003C= genomic_pos \u003C target_codon_end:\n                    target_position_in_spacer.append(i)\n            bystander_positions = [i for i in edit_bases_in_window if i not in target_position_in_spacer]\n            candidates.append({\n                'spacer': spacer,\n                'strand': strand,\n                'spacer_start': spacer_start,\n                'target_positions': target_position_in_spacer,\n                'bystander_positions': bystander_positions,\n                'n_bystanders': len(bystander_positions),\n            })\n    return pd.DataFrame(candidates).sort_values('n_bystanders')\n```\n\n**Decision rule:** Select spacers with target_positions != empty AND n_bystanders minimized. For variant-by-variant scanning, accept up to 1-2 bystanders if biology of those positions is interpretable; flag for downstream variant attribution.\n\n## Editing Efficiency Filtering (Critical Pre-Hit-Calling)\n\n**Goal:** Drop sgRNAs that do not edit efficiently, since unedited reads represent no biological perturbation.\n\n**Approach:** From CRISPResso2 output, compute target-base-conversion percentage per sgRNA; filter library to sgRNAs with >50% target editing in a pilot or co-screened control.\n\n```python\ndef filter_by_editing_efficiency(crispresso_outputs_dir, target_pos, target_base, efficiency_threshold=0.5):\n    '''Drop sgRNAs that edit \u003Cefficiency_threshold of reads at target position.\n    crispresso_outputs_dir: directory containing CRISPResso per-sample outputs.'''\n    from pathlib import Path\n    results = []\n    for sample_dir in Path(crispresso_outputs_dir).glob('CRISPResso_on_*'):\n        sgrna_id = sample_dir.name.replace('CRISPResso_on_', '')\n        quant_file = sample_dir \u002F 'Quantification_window_nucleotide_percentage_table.txt'\n        if not quant_file.exists():\n            continue\n        df = pd.read_csv(quant_file, sep='\\t')\n        # Find target position in the quantification window\n        target_row = df[df['Position'] == target_pos]\n        if target_row.empty:\n            continue\n        # Editing = sum of non-original bases at target position\n        original_pct = target_row[target_base].values[0]\n        editing_pct = (100 - original_pct) \u002F 100\n        results.append({'sgrna_id': sgrna_id, 'editing_pct': editing_pct,\n                         'pass_filter': editing_pct >= efficiency_threshold})\n    return pd.DataFrame(results)\n```\n\n**Convention:** Drop sgRNAs below 50% editing for variant-function screens. A common working split is a 30% editing floor for primary screening and a 50% floor for confirmed hits. Below 30%, the screen has insufficient power; above 70%, results approach saturation editing.\n\n## Bystander Edit Attribution\n\n**Why this matters:** When a sgRNA's editing window contains the target base AND a bystander base, the screen scores the combination. To attribute screen signal to the target variant alone, either (a) include sgRNAs that edit only the target (no bystander) -- often impossible -- or (b) deconvolute via parallel measurements.\n\n**Strategies for variant-by-variant attribution:**\n\n1. **Tile multiple sgRNAs with different bystander patterns:** If 5 different sgRNAs all hit the target base but have different bystanders, common signal across them is target-attributable (Hanna 2021 approach).\n\n2. **Use orthogonal chemistry:** Run the same variant scan with prime editor (no bystanders); cross-validate. See [[prime-editing-screens]].\n\n3. **Bystander stratification:** From CRISPResso2 allele table, partition reads by exact edit pattern (target only, target+bystander_1, target+bystander_2, etc.); separately score each pattern's contribution to the phenotype.\n\n4. **Restrict library:** Use only sgRNAs with zero bystanders in the editing window (rare; may exclude most candidate spacers).\n\n```python\ndef deconvolute_bystander(allele_table_path, target_pos, bystander_pos_list):\n    '''From CRISPResso2 allele table, partition reads by edit pattern at target + bystanders.\n    Returns: per-pattern frequency for each combination of target\u002Fbystander edits.'''\n    alleles = pd.read_csv(allele_table_path, sep='\\t', compression='zip')\n    # Mark target_edited and per-bystander_edited\n    alleles['target_edited'] = alleles['Aligned_Sequence'].str[target_pos-1] != alleles['Reference_Sequence'].str[target_pos-1]\n    for bp in bystander_pos_list:\n        alleles[f'bystander_{bp}_edited'] = alleles['Aligned_Sequence'].str[bp-1] != alleles['Reference_Sequence'].str[bp-1]\n    return alleles.groupby(['target_edited'] + [f'bystander_{bp}_edited' for bp in bystander_pos_list])['Reference_pct'].sum().reset_index()\n```\n\n## Hit Calling for Variant-Function Screens\n\n**Goal:** Score per-variant fitness from a base-editor screen.\n\n**Approach:** Filter library to efficiency-passing sgRNAs (>50% editing), then run MAGeCK MLE or drugZ on the sgRNA-level counts; map each significant sgRNA to its predicted variant + bystander pattern; aggregate to per-variant scores.\n\n```python\ndef aggregate_variant_scores(mageck_sgrna_summary, variant_annotation_df):\n    '''Aggregate sgRNA-level scores to per-variant scores.\n    variant_annotation_df: per-sgRNA -> predicted variants (target + bystanders).'''\n    df = mageck_sgrna_summary.merge(variant_annotation_df, on='sgRNA')\n    # Target-only contribution: sgRNAs with no bystanders\n    target_only = df[df['n_bystanders'] == 0]\n    target_only_scores = target_only.groupby('target_variant')['LFC'].agg(['mean', 'std', 'count'])\n    # Mixed signal: sgRNAs with bystanders\n    mixed = df[df['n_bystanders'] > 0]\n    return target_only_scores, mixed\n```\n\n## Hanna 2021 BRCA1\u002F2 Variant-Function Screen Methodology\n\n**Hanna et al 2021 *Cell* 184:1064** benchmarked CBE variant scanning at scale, screening 68,526 sgRNAs covering 52,034 ClinVar variants across 3,584 genes, with BRCA1 and BRCA2 as the positive\u002Fnegative-selection benchmark:\n\n1. Design the CBE library from predicted variant impact (ClinVar annotation), covering each variant with the sgRNAs that install it\n2. Run drug-modifier screens (PARPi sensitivity) with vehicle vs drug\n3. Score per variant by aggregating over all sgRNAs that install it; cross-check against bystander-controlled sgRNAs\n\n**Standard surrounding practice:** verify editing efficiency at a control timepoint via amplicon sequencing, drop low-efficiency sgRNAs (see the editing-efficiency convention above), and call sensitizers with a bidirectional method such as drugZ.\n\n**Quantified result:** Recovered known loss-of-function variants in BRCA1 and BRCA2 with high precision, and identified PARP1 variants conferring resistance to PARP inhibitors.\n\n## Cuella-Martin 2021 DDR-Gene Variant Screening\n\n**Cuella-Martin et al 2021 *Cell* 184:1081-1097** screened ~86 DNA-damage-response (DDR) genes (including BRCA1\u002F2) with CBE saturation mutagenesis:\n\n- Saturation CBE design across 86 DDR genes (not BRCA1\u002F2 alone)\n- Identified pathogenic\u002Flikely-pathogenic variants in critical protein domains\n- Combined with biochemical and genetic validation (for example the 53BP1-USP28 interaction surface)\n- Demonstrated saturation mutagenesis is feasible at protein-domain scale\n\n**Relationship to Hanna 2021:** the two studies appeared back-to-back in the same *Cell* issue and apply the same CBE variant-scanning strategy to complementary targets -- Hanna benchmarks against ClinVar-annotated variants genome-wide, Cuella-Martin saturates 86 DDR genes. Treat them as complementary methodology references, not as cross-validations of each other.\n\n## Cas9 vs Base Editor vs Prime Editor for Variant Installation\n\n| Approach | What it does | Bystander | Indels | When to use |\n|----------|--------------|-----------|--------|-------------|\n| Cas9 + HDR template | Installs precise edit + template | None | High (NHEJ competition) | When precise edit needed; high indel byproduct |\n| Cas9 (no template) | Random indels at cut site | None | 70%+ | Loss-of-function; not variant-specific |\n| CBE (BE3\u002FBE4) | C->T at editing window | Yes (multiple Cs) | \u003C5% | C->T variants with manageable bystanders |\n| ABE (ABE7.10\u002FABE8e) | A->G at editing window | Yes (multiple As) | \u003C2% | A->G variants; clean for single-A spacers |\n| CGBE \u002F GBE | C->G or C->A | Yes | 5-10% | Transversions; rare use cases |\n| Prime editor (PE2\u002FPE3) | Templated edit; any base change | None | 1-3% | Precise variants; lower efficiency |\n\n**Decision:** For C->T or A->G with available editing window: base editor is preferred (higher efficiency than PE). For other transitions\u002Ftransversions, multi-base edits, or zero-bystander requirements: prime editor.\n\n## Broad be-validation-pipeline\n\nThe Broad Institute's `be-validation-pipeline` (https:\u002F\u002Fbroadinstitute.github.io\u002Fbe-validation-pipeline\u002F) is a CRISPResso2 post-processing and validation toolkit for BE amplicon data -- a set of Jupyter notebooks, not a workflow-engine pipeline. Run CRISPResso2 first, then execute the notebooks in order:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbroadinstitute\u002Fbe-validation-pipeline\ncd be-validation-pipeline\npip install -r requirements.txt\n\n# Step 1: run CRISPResso2 in batch mode (or use the BEV tool on GPP LIMS).\n# The batch file is tab-delimited with columns: name, fastq_r1, amplicon_seq, guide_seq\n# (plus optional -w, -wc, --exclude_bp_from_left\u002Fright).\ndocker run -v ${PWD}:\u002FDATA -w \u002FDATA -i pinellolab\u002Fcrispresso2 \\\n    CRISPRessoBatch --batch_settings batch_file.txt --skip_failed --base_edit\n\n# Step 2: run the notebooks in order against the CRISPResso2 output\n#   notebooks\u002F01_BEV_allele_frequencies.ipynb\n#   notebooks\u002F02_BEV_nucleotide_percentage_plots.ipynb\n#   notebooks\u002F03_BEV_editing_efficiency.ipynb\n# Outputs: allele-frequency tables, nucleotide-percentage plots, editing-efficiency heat maps\n```\n\nThe notebooks cover allele-frequency tabulation, nucleotide-level editing quantification and editing-efficiency summaries. Hit calling is NOT part of this toolkit -- score the screen separately with drugZ or MAGeCK.\n\n## Failure Modes\n\n### Mostly indels in BE sample\n\n**Trigger:** Cas9 contamination, wrong vector (e.g., used pCas9-BE3 plasmid but selected on Cas9 line), or evoCDA-BE \u002F broader-window chemistry.\n**Mechanism:** Cas9 cuts dsDNA; BE relies on nicked-ssDNA deamination. Cas9 expression in the same cell creates indels.\n**Symptom:** Substitution-vs-indel ratio \u003C3 in CRISPResso output.\n**Fix:** Verify vector (nCas9-BE3 not Cas9-BE3); confirm cell line lacks Cas9 background; restrict to specifically engineered BE-cell lines.\n\n### High editing but no biological signal\n\n**Trigger:** Bystander C\u002FA is dominating; intended variant is not the perturbation driving phenotype.\n**Mechanism:** When target is at position 5 and bystander is at position 7, the molecule carries both; phenotype is from the bystander.\n**Symptom:** Strong screen signal but variant attribution unclear.\n**Fix:** Run orthogonal prime-editor scan of the same intended variants; restrict library to bystander-free spacers when possible; deconvolute via allele-frequency table.\n\n### sgRNA shows perfect editing but no fitness signal\n\n**Trigger:** Intended variant is silent or compensatory; the protein function is unchanged.\n**Mechanism:** Variants can be tolerated; not all variants are LoF or GoF.\n**Symptom:** High editing efficiency (>70%) but per-sgRNA LFC near zero.\n**Fix:** Expected outcome for many variants; flag silent \u002F compensatory variants in the report.\n\n### Low editing across all guides\n\n**Trigger:** Wrong cell line for the BE; cell line has poor BE activity (some lines lack APOBEC or have low expression).\n**Mechanism:** BE efficiency depends on cell-line expression of TadA or APOBEC components.\n**Symptom:** Median editing \u003C30% across library.\n**Fix:** Test in a BE-validated cell line (HEK293T, U2OS, K562 generally work); pilot before full screen.\n\n### Library missing intended-variant sgRNAs\n\n**Trigger:** No NGG-adjacent spacer places target base in editing window for that codon.\n**Mechanism:** Editor window is fixed; some codons cannot be targeted with given chemistry.\n**Symptom:** Specific variants absent from screen.\n**Fix:** Use PAM-relaxed BE variants (SpRY-CBE, SpRY-ABE); use prime editor for variants outside BE accessibility; accept that some variants cannot be installed.\n\n## Quantitative Thresholds\n\n| Threshold | Value | Source \u002F Rationale |\n|-----------|-------|--------------------|\n| Editing window | Positions 4-8 from PAM-distal end (BE3\u002FBE4); 4-7 (ABE7.10); 4-8 (SpABE8e) | Komor 2016; Gaudelli 2017; Richter 2020 |\n| Editing efficiency for screen power | >30% (primary); >50% (validation) | Field convention (BE variant screens) |\n| Indel byproduct (clean BE) | \u003C5%; \u003C2% for ABE | Koblan 2018 (BE4max); Gaudelli 2017 (ABE) |\n| Substitution-vs-indel ratio | >10 (clean BE); \u003C3 (Cas9-like) | CRISPResso2 diagnostic |\n| Bystander rate (target attribution) | \u003C10% acceptable; \u003C5% ideal for clean attribution | Application-dependent |\n| Cell-line BE activity (pilot) | >30% editing at validated target | Below = wrong cell line for BE |\n| Per-amino-acid sgRNA density | 10-15 (saturation designs); 5-8 (smaller screens) | Tradeoff with library size |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| Substitution-vs-indel ratio \u003C3 | Cas9 contamination or wrong BE | Verify vector \u002F cell line; pilot first |\n| All edits at bystander positions | Target base outside window | Re-design spacer with target at pos 4-8 |\n| Variant attribution unclear | Bystander confounding | Run orthogonal PE; restrict library |\n| Library lacks intended variant | No NGG-PAM accessibility | SpRY-CBE; prime editor; accept exclusion |\n| Editing \u003C30% library-wide | Cell-line BE inactivity | Re-validate cell line |\n| Hit list dominated by single sgRNA | Bystander-driven phenotype | Cross-check with bystander-free sgRNAs |\n\n## References\n\n- Komor AC et al. 2016. *Nature* 533:420. BE3.\n- Gaudelli NM et al. 2017. *Nature* 551:464. ABE7.10.\n- Koblan LW et al. 2018. *Nat Biotechnol* 36:843. BE4max + improved CBE.\n- Richter MF et al. 2020. *Nat Biotechnol* 38:883. ABE8e; phage-assisted evolution of ABE7.10.\n- Lapinaite A et al. 2020. *Science* 369:566. ABE8e mechanism.\n- Gaudelli NM et al. 2020. *Nat Biotechnol* 38:892. ABE8 series (ABE8.20).\n- Hanna RE et al. 2021. *Cell* 184:1064. Massively parallel BRCA1\u002F2 variant function via CBE.\n- Cuella-Martin R et al. 2021. *Cell* 184:1081-1097. CBE saturation across 86 DDR genes (BRCA1\u002F2 plus others).\n- Arbab M et al. 2020. *Cell* 182:463. BE-Hive prediction of editing outcomes.\n- Anzalone AV et al. 2019. *Nature* 576:149. Prime editing (PE2\u002FPE3).\n- Clement K et al. 2019. *Nat Biotechnol* 37:224. CRISPResso2.\n- Kurt IC et al. 2021. *Nat Biotechnol* 39:41. CGBE1.\n\n## Related Skills\n\n- crispr-screens\u002Fcrispresso-editing - CRISPResso2 BE\u002FPE mode and allele tables\n- crispr-screens\u002Flibrary-design - base-editor library design\n- crispr-screens\u002Fprime-editing-screens - Orthogonal PE for variant attribution\n- crispr-screens\u002Fhit-calling - Variant-level hit aggregation\n- crispr-screens\u002Fscreen-qc - Editing-efficiency QC\n- crispr-screens\u002Fdrugz-chemogenomic - drugZ for BE drug-modifier screens\n- clinical-databases\u002Fclinvar-lookup - Variant pathogenicity annotation\n- variant-calling\u002Fvariant-annotation - VEP for predicted amino acid changes\n","## Version Compatibility\n\nReference examples tested with: CRISPResso2 2.2.14+, BE-Hive 1.0+ (BE prediction), pandas 2.2+, biopython 1.83+, numpy 1.26+, scipy 1.12+, scikit-learn 1.4+; Broad be-validation-pipeline notebooks (repo HEAD).\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `CRISPResso --version`\n- Python: `pip show CRISPResso2`; BE-Hive is a GitHub clone (maxwshen\u002Fbe_predict_bystander), not a PyPI package\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.\n\n## Base Editing Screen Analysis\n\n**\"Analyze my base-editor variant-function screen\"** -> Quantify per-sgRNA target-base conversion, bystander rate, and indel byproducts from amplicon sequencing; filter on editing efficiency; map each sgRNA to its intended SNV (target + bystander pattern); compute per-variant fitness from the screen log-fold change; reconcile target vs bystander variant attribution; annotate against ClinVar \u002F COSMIC.\n\n- CLI: `CRISPResso --base_editor_output` for per-amplicon BE quantification\n- CLI: Broad `be-validation-pipeline` for end-to-end pooled-screen analysis with editing-efficiency filtering\n- Python: `BE-Hive` (Arbab 2020) for editing-efficiency prediction; clone maxwshen\u002Fbe_predict_bystander and import via sys.path\n- Web: `BE-Designer` (Hwang 2018, RGEN Tools) for variant-encoding sgRNA design\n\n## Base Editor Chemistry Selection\n\n| Editor | Reaction | Editing window | Indel byproduct rate | When to use |\n|--------|----------|----------------|----------------------|-------------|\n| BE3 (Komor 2016) | C->T (also G->A on opposite strand) | Pos 4-8 from PAM-distal end | 5-10% | Original; superseded |\n| BE4 \u002F BE4max (Koblan 2018) | C->T | Pos 4-8 | \u003C5% | CBE standard |\n| eA3A-BE3 | C->T narrow specificity | Pos 5-7 | \u003C5% | Specifically TC contexts (eA3A prefers TC) |\n| ABE7.10 (Gaudelli 2017) | A->G (T->C opposite strand) | Pos 4-7 | \u003C2% | First ABE; slow at non-TA contexts |\n| ABE8.20 (Gaudelli 2020) | A->G | Pos 4-8 | \u003C2% | Modern ABE; high activity |\n| ABE8e (Richter 2020) | A->G | Pos 4-8 | \u003C2% | Highest editing activity; more processive than ABE7.10 |\n| evoCDA-BE | C->T (broader) | Pos 1-9 | 5-10% | Larger editing window; more bystander |\n| CGBE1 (Kurt 2021) | C->G | Pos 5-7 | 5-10% | C-to-G transversion; rare use |\n| GBE (Zhao 2021) | C->G or C->A | Pos 4-7 | 5-10% | Transversions; less mature |\n\n**Decision rule:** For a target SNV at position 4-8 of a candidate spacer with no bystander Cs\u002FAs in the same window, BE3-BE4 or ABE7.10 is sufficient. For high-throughput variant scanning where bystander tolerance must be minimized, use eA3A-BE3 (TC contexts only) for C->T, or ABE7.10 rather than ABE8e\u002FABE8.20 for A->G -- its 4-7 window is the narrowest ABE.\n\n## Editing Window Math\n\n**Why this matters for postdoc-level use:** Base editors are tethered to dCas9 (or nCas9) and the deaminase acts on the displaced ssDNA \"R-loop\" formed when Cas9 binds. The deaminase has a fixed reach -- positions 4-8 from the PAM-distal end of the protospacer for canonical BE3\u002FBE4, and 4-7 for ABE7.10. Outside this window, editing efficiency drops by 10-50x.\n\n```\nPAM-distal end                                                            PAM-proximal\n   |                                                                          |\n   1  2  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 20    NGG\n                  ^^^^^^^^^^^\n                  Canonical editing window (positions 4-8)\n\n   For BE4max: positions 4-8 are 5-50x more efficient than positions 1-3 or 9-13 (ABE7.10: 4-7)\n   For SpABE8e: positions 4-8 (Richter 2020), matching the corresponding CBEs rather than ABE7.10's narrower 4-7\n   For evoCDA-BE: window 1-9 (broader; more bystander)\n```\n\n**Critical implication for variant interpretation:** If the intended edit is at position 5 and there is an additional editable C\u002FA at position 7, both will be edited in the same molecule. The screen scores the *combination* of edits, not the intended one alone. This is bystander confounding.\n\n## sgRNA Library Design for BE Screens\n\n**Goal:** Tile editing-window-positioned spacers across a protein region of interest to enable variant scanning.\n\n**Approach:** For each amino acid in the target region, find NGG-adjacent spacers where the SNV-of-interest base falls in editing positions 4-8 with minimal bystander C\u002FA in the same window. Annotate each spacer with the predicted amino acid changes (target + bystander).\n\n```python\nimport pandas as pd\nimport re\nfrom Bio.Seq import Seq\n\ndef find_be_spacers(cds_sequence, cds_protein_start, target_aa, target_base='C', editor='BE4max'):\n    '''Find sgRNAs that place target_base in editor-specific window at target_aa.\n    Returns spacers with bystander annotation.\n\n    Args:\n        cds_sequence: nucleotide CDS (translated frame 1)\n        cds_protein_start: amino acid number of CDS start (usually 1)\n        target_aa: amino acid number to install variant (e.g., 130 for residue 130)\n        target_base: 'C' (CBE) or 'A' (ABE)\n        editor: 'BE3', 'BE4max', 'eA3A-BE3', 'ABE7.10', 'ABE8.20', 'ABE8e', 'evoCDA-BE'\n\n    Returns: DataFrame with spacer, position-in-cds, target-base-position-in-spacer,\n             bystander_positions, predicted_aa_changes\n    '''\n    # Editor-specific editing window (positions from PAM-distal end of spacer)\n    window_by_editor = {\n        'BE3': (4, 8),       'BE4max': (4, 8),    'eA3A-BE3': (5, 7),\n        'ABE7.10': (4, 7),   'ABE8.20': (4, 8),   'ABE8e': (4, 8),     # SpABE8e matches CBE window (Richter 2020)\n        'evoCDA-BE': (1, 9),\n    }\n    window_lo, window_hi = window_by_editor[editor]\n    aa_index = target_aa - cds_protein_start  # 0-indexed in protein\n    aa_start_nt = aa_index * 3                # nt offset in cds\n    candidates = []\n    spacer_len = 20\n    pam_pattern = re.compile(r'(?=([ACGT]GG))')\n    for strand, seq in [('+', cds_sequence), ('-', str(Seq(cds_sequence).reverse_complement()))]:\n        for pam_match in pam_pattern.finditer(seq):\n            pam_pos = pam_match.start()\n            spacer_start = pam_pos - spacer_len\n            if spacer_start \u003C 0:\n                continue\n            spacer = seq[spacer_start:pam_pos]\n            # Editor-specific window from PAM-distal end (1-indexed)\n            # Find all editable bases in window\n            edit_bases_in_window = []\n            for i, b in enumerate(spacer[window_lo-1:window_hi], start=window_lo):\n                if b == target_base:\n                    edit_bases_in_window.append(i)\n            if not edit_bases_in_window:\n                continue\n            # Annotate which edits hit the target_aa codon\n            target_codon_start = aa_start_nt\n            target_codon_end = target_codon_start + 3\n            target_position_in_spacer = []\n            for i in edit_bases_in_window:\n                genomic_pos = spacer_start + i - 1\n                if target_codon_start \u003C= genomic_pos \u003C target_codon_end:\n                    target_position_in_spacer.append(i)\n            bystander_positions = [i for i in edit_bases_in_window if i not in target_position_in_spacer]\n            candidates.append({\n                'spacer': spacer,\n                'strand': strand,\n                'spacer_start': spacer_start,\n                'target_positions': target_position_in_spacer,\n                'bystander_positions': bystander_positions,\n                'n_bystanders': len(bystander_positions),\n            })\n    return pd.DataFrame(candidates).sort_values('n_bystanders')\n```\n\n**Decision rule:** Select spacers with target_positions != empty AND n_bystanders minimized. For variant-by-variant scanning, accept up to 1-2 bystanders if biology of those positions is interpretable; flag for downstream variant attribution.\n\n## Editing Efficiency Filtering (Critical Pre-Hit-Calling)\n\n**Goal:** Drop sgRNAs that do not edit efficiently, since unedited reads represent no biological perturbation.\n\n**Approach:** From CRISPResso2 output, compute target-base-conversion percentage per sgRNA; filter library to sgRNAs with >50% target editing in a pilot or co-screened control.\n\n```python\ndef filter_by_editing_efficiency(crispresso_outputs_dir, target_pos, target_base, efficiency_threshold=0.5):\n    '''Drop sgRNAs that edit \u003Cefficiency_threshold of reads at target position.\n    crispresso_outputs_dir: directory containing CRISPResso per-sample outputs.'''\n    from pathlib import Path\n    results = []\n    for sample_dir in Path(crispresso_outputs_dir).glob('CRISPResso_on_*'):\n        sgrna_id = sample_dir.name.replace('CRISPResso_on_', '')\n        quant_file = sample_dir \u002F 'Quantification_window_nucleotide_percentage_table.txt'\n        if not quant_file.exists():\n            continue\n        df = pd.read_csv(quant_file, sep='\\t')\n        # Find target position in the quantification window\n        target_row = df[df['Position'] == target_pos]\n        if target_row.empty:\n            continue\n        # Editing = sum of non-original bases at target position\n        original_pct = target_row[target_base].values[0]\n        editing_pct = (100 - original_pct) \u002F 100\n        results.append({'sgrna_id': sgrna_id, 'editing_pct': editing_pct,\n                         'pass_filter': editing_pct >= efficiency_threshold})\n    return pd.DataFrame(results)\n```\n\n**Convention:** Drop sgRNAs below 50% editing for variant-function screens. A common working split is a 30% editing floor for primary screening and a 50% floor for confirmed hits. Below 30%, the screen has insufficient power; above 70%, results approach saturation editing.\n\n## Bystander Edit Attribution\n\n**Why this matters:** When a sgRNA's editing window contains the target base AND a bystander base, the screen scores the combination. To attribute screen signal to the target variant alone, either (a) include sgRNAs that edit only the target (no bystander) -- often impossible -- or (b) deconvolute via parallel measurements.\n\n**Strategies for variant-by-variant attribution:**\n\n1. **Tile multiple sgRNAs with different bystander patterns:** If 5 different sgRNAs all hit the target base but have different bystanders, common signal across them is target-attributable (Hanna 2021 approach).\n\n2. **Use orthogonal chemistry:** Run the same variant scan with prime editor (no bystanders); cross-validate. See [[prime-editing-screens]].\n\n3. **Bystander stratification:** From CRISPResso2 allele table, partition reads by exact edit pattern (target only, target+bystander_1, target+bystander_2, etc.); separately score each pattern's contribution to the phenotype.\n\n4. **Restrict library:** Use only sgRNAs with zero bystanders in the editing window (rare; may exclude most candidate spacers).\n\n```python\ndef deconvolute_bystander(allele_table_path, target_pos, bystander_pos_list):\n    '''From CRISPResso2 allele table, partition reads by edit pattern at target + bystanders.\n    Returns: per-pattern frequency for each combination of target\u002Fbystander edits.'''\n    alleles = pd.read_csv(allele_table_path, sep='\\t', compression='zip')\n    # Mark target_edited and per-bystander_edited\n    alleles['target_edited'] = alleles['Aligned_Sequence'].str[target_pos-1] != alleles['Reference_Sequence'].str[target_pos-1]\n    for bp in bystander_pos_list:\n        alleles[f'bystander_{bp}_edited'] = alleles['Aligned_Sequence'].str[bp-1] != alleles['Reference_Sequence'].str[bp-1]\n    return alleles.groupby(['target_edited'] + [f'bystander_{bp}_edited' for bp in bystander_pos_list])['Reference_pct'].sum().reset_index()\n```\n\n## Hit Calling for Variant-Function Screens\n\n**Goal:** Score per-variant fitness from a base-editor screen.\n\n**Approach:** Filter library to efficiency-passing sgRNAs (>50% editing), then run MAGeCK MLE or drugZ on the sgRNA-level counts; map each significant sgRNA to its predicted variant + bystander pattern; aggregate to per-variant scores.\n\n```python\ndef aggregate_variant_scores(mageck_sgrna_summary, variant_annotation_df):\n    '''Aggregate sgRNA-level scores to per-variant scores.\n    variant_annotation_df: per-sgRNA -> predicted variants (target + bystanders).'''\n    df = mageck_sgrna_summary.merge(variant_annotation_df, on='sgRNA')\n    # Target-only contribution: sgRNAs with no bystanders\n    target_only = df[df['n_bystanders'] == 0]\n    target_only_scores = target_only.groupby('target_variant')['LFC'].agg(['mean', 'std', 'count'])\n    # Mixed signal: sgRNAs with bystanders\n    mixed = df[df['n_bystanders'] > 0]\n    return target_only_scores, mixed\n```\n\n## Hanna 2021 BRCA1\u002F2 Variant-Function Screen Methodology\n\n**Hanna et al 2021 *Cell* 184:1064** benchmarked CBE variant scanning at scale, screening 68,526 sgRNAs covering 52,034 ClinVar variants across 3,584 genes, with BRCA1 and BRCA2 as the positive\u002Fnegative-selection benchmark:\n\n1. Design the CBE library from predicted variant impact (ClinVar annotation), covering each variant with the sgRNAs that install it\n2. Run drug-modifier screens (PARPi sensitivity) with vehicle vs drug\n3. Score per variant by aggregating over all sgRNAs that install it; cross-check against bystander-controlled sgRNAs\n\n**Standard surrounding practice:** verify editing efficiency at a control timepoint via amplicon sequencing, drop low-efficiency sgRNAs (see the editing-efficiency convention above), and call sensitizers with a bidirectional method such as drugZ.\n\n**Quantified result:** Recovered known loss-of-function variants in BRCA1 and BRCA2 with high precision, and identified PARP1 variants conferring resistance to PARP inhibitors.\n\n## Cuella-Martin 2021 DDR-Gene Variant Screening\n\n**Cuella-Martin et al 2021 *Cell* 184:1081-1097** screened ~86 DNA-damage-response (DDR) genes (including BRCA1\u002F2) with CBE saturation mutagenesis:\n\n- Saturation CBE design across 86 DDR genes (not BRCA1\u002F2 alone)\n- Identified pathogenic\u002Flikely-pathogenic variants in critical protein domains\n- Combined with biochemical and genetic validation (for example the 53BP1-USP28 interaction surface)\n- Demonstrated saturation mutagenesis is feasible at protein-domain scale\n\n**Relationship to Hanna 2021:** the two studies appeared back-to-back in the same *Cell* issue and apply the same CBE variant-scanning strategy to complementary targets -- Hanna benchmarks against ClinVar-annotated variants genome-wide, Cuella-Martin saturates 86 DDR genes. Treat them as complementary methodology references, not as cross-validations of each other.\n\n## Cas9 vs Base Editor vs Prime Editor for Variant Installation\n\n| Approach | What it does | Bystander | Indels | When to use |\n|----------|--------------|-----------|--------|-------------|\n| Cas9 + HDR template | Installs precise edit + template | None | High (NHEJ competition) | When precise edit needed; high indel byproduct |\n| Cas9 (no template) | Random indels at cut site | None | 70%+ | Loss-of-function; not variant-specific |\n| CBE (BE3\u002FBE4) | C->T at editing window | Yes (multiple Cs) | \u003C5% | C->T variants with manageable bystanders |\n| ABE (ABE7.10\u002FABE8e) | A->G at editing window | Yes (multiple As) | \u003C2% | A->G variants; clean for single-A spacers |\n| CGBE \u002F GBE | C->G or C->A | Yes | 5-10% | Transversions; rare use cases |\n| Prime editor (PE2\u002FPE3) | Templated edit; any base change | None | 1-3% | Precise variants; lower efficiency |\n\n**Decision:** For C->T or A->G with available editing window: base editor is preferred (higher efficiency than PE). For other transitions\u002Ftransversions, multi-base edits, or zero-bystander requirements: prime editor.\n\n## Broad be-validation-pipeline\n\nThe Broad Institute's `be-validation-pipeline` (https:\u002F\u002Fbroadinstitute.github.io\u002Fbe-validation-pipeline\u002F) is a CRISPResso2 post-processing and validation toolkit for BE amplicon data -- a set of Jupyter notebooks, not a workflow-engine pipeline. Run CRISPResso2 first, then execute the notebooks in order:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbroadinstitute\u002Fbe-validation-pipeline\ncd be-validation-pipeline\npip install -r requirements.txt\n\n# Step 1: run CRISPResso2 in batch mode (or use the BEV tool on GPP LIMS).\n# The batch file is tab-delimited with columns: name, fastq_r1, amplicon_seq, guide_seq\n# (plus optional -w, -wc, --exclude_bp_from_left\u002Fright).\ndocker run -v ${PWD}:\u002FDATA -w \u002FDATA -i pinellolab\u002Fcrispresso2 \\\n    CRISPRessoBatch --batch_settings batch_file.txt --skip_failed --base_edit\n\n# Step 2: run the notebooks in order against the CRISPResso2 output\n#   notebooks\u002F01_BEV_allele_frequencies.ipynb\n#   notebooks\u002F02_BEV_nucleotide_percentage_plots.ipynb\n#   notebooks\u002F03_BEV_editing_efficiency.ipynb\n# Outputs: allele-frequency tables, nucleotide-percentage plots, editing-efficiency heat maps\n```\n\nThe notebooks cover allele-frequency tabulation, nucleotide-level editing quantification and editing-efficiency summaries. Hit calling is NOT part of this toolkit -- score the screen separately with drugZ or MAGeCK.\n\n## Failure Modes\n\n### Mostly indels in BE sample\n\n**Trigger:** Cas9 contamination, wrong vector (e.g., used pCas9-BE3 plasmid but selected on Cas9 line), or evoCDA-BE \u002F broader-window chemistry.\n**Mechanism:** Cas9 cuts dsDNA; BE relies on nicked-ssDNA deamination. Cas9 expression in the same cell creates indels.\n**Symptom:** Substitution-vs-indel ratio \u003C3 in CRISPResso output.\n**Fix:** Verify vector (nCas9-BE3 not Cas9-BE3); confirm cell line lacks Cas9 background; restrict to specifically engineered BE-cell lines.\n\n### High editing but no biological signal\n\n**Trigger:** Bystander C\u002FA is dominating; intended variant is not the perturbation driving phenotype.\n**Mechanism:** When target is at position 5 and bystander is at position 7, the molecule carries both; phenotype is from the bystander.\n**Symptom:** Strong screen signal but variant attribution unclear.\n**Fix:** Run orthogonal prime-editor scan of the same intended variants; restrict library to bystander-free spacers when possible; deconvolute via allele-frequency table.\n\n### sgRNA shows perfect editing but no fitness signal\n\n**Trigger:** Intended variant is silent or compensatory; the protein function is unchanged.\n**Mechanism:** Variants can be tolerated; not all variants are LoF or GoF.\n**Symptom:** High editing efficiency (>70%) but per-sgRNA LFC near zero.\n**Fix:** Expected outcome for many variants; flag silent \u002F compensatory variants in the report.\n\n### Low editing across all guides\n\n**Trigger:** Wrong cell line for the BE; cell line has poor BE activity (some lines lack APOBEC or have low expression).\n**Mechanism:** BE efficiency depends on cell-line expression of TadA or APOBEC components.\n**Symptom:** Median editing \u003C30% across library.\n**Fix:** Test in a BE-validated cell line (HEK293T, U2OS, K562 generally work); pilot before full screen.\n\n### Library missing intended-variant sgRNAs\n\n**Trigger:** No NGG-adjacent spacer places target base in editing window for that codon.\n**Mechanism:** Editor window is fixed; some codons cannot be targeted with given chemistry.\n**Symptom:** Specific variants absent from screen.\n**Fix:** Use PAM-relaxed BE variants (SpRY-CBE, SpRY-ABE); use prime editor for variants outside BE accessibility; accept that some variants cannot be installed.\n\n## Quantitative Thresholds\n\n| Threshold | Value | Source \u002F Rationale |\n|-----------|-------|--------------------|\n| Editing window | Positions 4-8 from PAM-distal end (BE3\u002FBE4); 4-7 (ABE7.10); 4-8 (SpABE8e) | Komor 2016; Gaudelli 2017; Richter 2020 |\n| Editing efficiency for screen power | >30% (primary); >50% (validation) | Field convention (BE variant screens) |\n| Indel byproduct (clean BE) | \u003C5%; \u003C2% for ABE | Koblan 2018 (BE4max); Gaudelli 2017 (ABE) |\n| Substitution-vs-indel ratio | >10 (clean BE); \u003C3 (Cas9-like) | CRISPResso2 diagnostic |\n| Bystander rate (target attribution) | \u003C10% acceptable; \u003C5% ideal for clean attribution | Application-dependent |\n| Cell-line BE activity (pilot) | >30% editing at validated target | Below = wrong cell line for BE |\n| Per-amino-acid sgRNA density | 10-15 (saturation designs); 5-8 (smaller screens) | Tradeoff with library size |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| Substitution-vs-indel ratio \u003C3 | Cas9 contamination or wrong BE | Verify vector \u002F cell line; pilot first |\n| All edits at bystander positions | Target base outside window | Re-design spacer with target at pos 4-8 |\n| Variant attribution unclear | Bystander confounding | Run orthogonal PE; restrict library |\n| Library lacks intended variant | No NGG-PAM accessibility | SpRY-CBE; prime editor; accept exclusion |\n| Editing \u003C30% library-wide | Cell-line BE inactivity | Re-validate cell line |\n| Hit list dominated by single sgRNA | Bystander-driven phenotype | Cross-check with bystander-free sgRNAs |\n\n## References\n\n- Komor AC et al. 2016. *Nature* 533:420. BE3.\n- Gaudelli NM et al. 2017. *Nature* 551:464. ABE7.10.\n- Koblan LW et al. 2018. *Nat Biotechnol* 36:843. BE4max + improved CBE.\n- Richter MF et al. 2020. *Nat Biotechnol* 38:883. ABE8e; phage-assisted evolution of ABE7.10.\n- Lapinaite A et al. 2020. *Science* 369:566. ABE8e mechanism.\n- Gaudelli NM et al. 2020. *Nat Biotechnol* 38:892. ABE8 series (ABE8.20).\n- Hanna RE et al. 2021. *Cell* 184:1064. Massively parallel BRCA1\u002F2 variant function via CBE.\n- Cuella-Martin R et al. 2021. *Cell* 184:1081-1097. CBE saturation across 86 DDR genes (BRCA1\u002F2 plus others).\n- Arbab M et al. 2020. *Cell* 182:463. BE-Hive prediction of editing outcomes.\n- Anzalone AV et al. 2019. *Nature* 576:149. Prime editing (PE2\u002FPE3).\n- Clement K et al. 2019. *Nat Biotechnol* 37:224. CRISPResso2.\n- Kurt IC et al. 2021. *Nat Biotechnol* 39:41. CGBE1.\n\n## Related Skills\n\n- crispr-screens\u002Fcrispresso-editing - CRISPResso2 BE\u002FPE mode and allele tables\n- crispr-screens\u002Flibrary-design - base-editor library design\n- crispr-screens\u002Fprime-editing-screens - Orthogonal PE for variant attribution\n- crispr-screens\u002Fhit-calling - Variant-level hit aggregation\n- crispr-screens\u002Fscreen-qc - Editing-efficiency QC\n- crispr-screens\u002Fdrugz-chemogenomic - drugZ for BE drug-modifier screens\n- clinical-databases\u002Fclinvar-lookup - Variant pathogenicity annotation\n- variant-calling\u002Fvariant-annotation - VEP for predicted amino acid changes\n",{"name":5,"source":24,"license":15,"tool_type":25,"description":7,"imported_at":26,"source_path":27,"primary_tool":28,"source_repository":29},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills","mixed","2026-09-20","crispr-screens\u002Fbase-editing-analysis\u002FSKILL.md","CRISPResso2","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-crispr-screens-base-editing-analysis","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-crispr-screens-base-editing-analysis",{"tool":42,"command":43},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-crispr-screens-base-editing-analysis",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-crispr-screens-base-editing-analysis","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-crispr-screens-base-editing-analysis\u002FSKILL.md",true,[50,56],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"examples\u002Fbase_editing_analysis.sh","SCRIPT","脚本",4119,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-crispr-screens-base-editing-analysis\u002Fexamples\u002Fbase_editing_analysis.sh",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"usage-guide.md","REFERENCE","参考文档",9075,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-crispr-screens-base-editing-analysis\u002Fusage-guide.md",[63,71,78,85,92,99],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":70},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":70},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":70},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":70},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":70},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":70},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]