bio-crispr-screens-library-design

## Version Compatibility Reference examples tested with: CRISPOR 5.01+, BioPython 1.83+, pandas 2.2+, numpy 1.26+, Azimuth 2.0+ (Doench 2016), CRISPRon 1.0+ (Xiang 2021), DeepSpCas9 1.0+ (Kim 2019). Before using code patterns, verify installed versions match. If versions differ: - CLI: `crispor.py --help` from the crisporWebsite clone - Python: Azimuth has no console script; call `azimuth.model_comparison.predict(...)` If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying. ## sgRNA Library Design **"Design a CRISPR library for my screen"** -> Pick a chemistry (Cas9 KO, CRISPRi, CRISPRa, Cas12a, base or prime editor), score candidate guides for on-target activity and off-target liability, position them relative to gene/TSS, add appropriate controls, lay out the oligo for synthesis, and validate the cloned pool. - Python: `crispor.py` (web + CLI) for batch genome-wide guide scoring with CFD+MIT off-target - Python: `azimuth` (Microsoft Research) for Rule Set 2 on-target predictions (Brunello-style) - Python: `CRISPRon`, `DeepSpCas9` for modern deep-learning predictors - R: `crisprDesign` (Bioconductor) for integrated annotation-aware design ## Library Chemistry Decision Tree | Goal | Chemistry | Canonical library | Guides/gene | TSS / target window | |------|-----------|-------------------|-------------|---------------------| | Loss-of-function essentiality, fitness | SpCas9 KO | Brunello, TKOv3, Avana | 4 (Brunello), 4 (TKOv3), 6 (Avana) | Constitutive exons, prefer aa 5-65% from N-terminus | | Knockdown of non-cuttable genes, dosage-sensitive | dCas9-KRAB (CRISPRi) | Dolcetto, Horlbeck v2 | 6 (Dolcetto), 5 (Horlbeck) | Optimum +25 to +75 downstream of FANTOM5 TSS, searched out to -50/+300 (Dolcetto, Sanson 2018); -25 to +500 (Horlbeck v2) | | Gain-of-function, gene activation | dCas9-VP64 / SAM / SunTag (CRISPRa) | Calabrese, Horlbeck-CRISPRa | 6 (Calabrese), 5 (Horlbeck) | -150 to -75 from TSS (Calabrese); -550 to -25 (Horlbeck v2) | | Paralog buffering, GI screens | enAsCas12a multiplex | Inzolia, in4mer | 4-guide arrays | Constitutive exons | | Variant function, SNV scanning | CBE / ABE | Custom tiling library | Tile editing windows | Editing window pos 4-8 from PAM-distal end | | Precise edit, indel-free | Prime editor | Custom PRIDICT-designed | Tile pegRNAs | Anywhere with NGG PAM within 30 nt of edit | **Fails when:** - CRISPRi/a targeting wrong TSS: any TSS without FANTOM5 CAGE evidence is suspect; guides positioned against the wrong TSS lose most of their knockdown. - Cas9 KO of essential paralogs: single-KO buffering hides paralog-redundant essentials (42% of constitutively expressed genes never score, Dede 2020); switch to Cas12a multiplex. - Base editor over an exon-intron boundary: editing-window bystanders create splice variants instead of the intended SNV. ## On-Target Scoring: Algorithmic Taxonomy | Predictor | Year | Training set | Strengths | Fails when | |-----------|------|--------------|-----------|------------| | Doench Rule Set 1 | 2014 | Flow-sorted GFP+ knockouts | Simple, interpretable | Limited training data; sub-optimal at >NGG context | | Doench Rule Set 2 / Azimuth 2.0 | 2016 | 1,841 flow-cytometry guides (Doench 2014) plus new guides tiling additional genes | Gold-standard for SpCas9; basis of Brunello | Trained on dropouts; under-predicts efficacy for nuclear-localized targets | | DeepSpCas9 | 2019 | 12,832 synthetic targets integrated in HEK293T | Spearman ~0.77 vs measured indel frequency on held-out data | Black-box; sensitive to chromatin/context features it wasn't trained on | | CRISPRon | 2021 | High-throughput indel sequencing | Best for therapeutic-grade target nomination | Slow per-guide; over-fits to its specific cell line | | DeepHF | 2019 | ~171k guides in HEK293T (WT 55,604; eSpCas9(1.1) 58,167; HF1 56,888) | Separate model per enzyme variant, including WT | Pick the model matching the enzyme actually used | **Reconciliation:** When predictors disagree, prefer the model whose training cell line matches the screen line (DeepSpCas9 was trained on synthetic targets integrated in HEK293T). For Brunello selection, Azimuth/Rule Set 2 is sufficient because the library was built with it -- introducing a different scorer creates apples-to-oranges ranking with the original library. ## Off-Target Scoring | Score | Year | Math | Cutoff convention | |-------|------|------|--------------------| | MIT (Hsu) | 2013 | Position-weighted mismatch penalty | Specificity score 0-100, higher is better; CRISPOR treats >=50 as a good guide | | CFD (Doench) | 2016 | Position+nucleotide-specific penalty fit on Brunello | Per-site CFD >0.2 counts a candidate off-target (Doench 2016); CRISPOR's aggregate CFD specificity score is 0-100, higher is better | | Elevation | 2018 | ML on CFD + mismatch positions | Tighter than CFD | CFD remains the default for genome-wide library design. **Critical pitfall:** CFD penalizes only mismatches, not bulges; for ≤1 mismatch + 1-bp bulge off-targets, validate empirically with GUIDE-seq or CIRCLE-seq. CRISPOR reports both the MIT (Hsu) and CFD guide specificity scores in a single output. ## Score and Rank sgRNAs for a Target Gene **Goal:** Generate ranked sgRNA candidates for a single gene, jointly scored on on-target activity (Rule Set 2 / Azimuth) and off-target liability (CFD). **Approach:** Identify all PAM-adjacent 20-nt protospacers in the target gene's coding sequence, retain only those in the first 5-65% of the protein (constitutive-exon convention from Brunello), filter on GC 30-70% and absence of poly-T (≥4 Ts terminates U6), call Azimuth for on-target and CRISPOR for off-target, and select the top N satisfying both criteria. ```python import re import pandas as pd import numpy as np from Bio.Seq import Seq def find_sgrna_candidates(cds_sequence, pam='NGG', guide_length=20): '''Return all protospacer candidates with PAM coordinates on + strand. Caller must filter by exon position and Azimuth/CFD score.''' pam_pattern = re.compile(f'(?=([ACGT]{{{guide_length}}}{pam.replace("N", "[ACGT]")}))') candidates = [] for strand, seq in [('+', cds_sequence), ('-', str(Seq(cds_sequence).reverse_complement()))]: for m in pam_pattern.finditer(seq): spacer = m.group(1)[:guide_length] if 'TTTT' in spacer or spacer.count('G') + spacer.count('C') not in range(6, 15): continue candidates.append({'spacer': spacer, 'strand': strand, 'pos_in_cds': m.start() if strand == '+' else len(seq) - m.start() - 23, 'gc_frac': (spacer.count('G') + spacer.count('C')) / guide_length}) return pd.DataFrame(candidates) def annotate_exon_position(candidates_df, cds_length): '''Filter to protospacers within first 5-65% of CDS (Brunello convention). Reason: N-terminal indels truncate protein; very-N-terminal hits alt initiation; C-terminal hits miss functional domains (Doench 2016 Nat Biotech).''' lo, hi = 0.05 * cds_length, 0.65 * cds_length return candidates_df[(candidates_df['pos_in_cds'] >= lo) & (candidates_df['pos_in_cds'] <= hi)].copy() ``` ## CRISPRi / CRISPRa TSS Targeting **Goal:** Position guides relative to the empirical TSS for maximum knockdown (CRISPRi) or activation (CRISPRa). **Approach:** Resolve TSS from FANTOM5 CAGE peaks (highest-ranked peak per gene; fall back to Ensembl/RefSeq if absent), define the modality-specific window, score candidate spacers in that window with Rule Set 2 plus the Horlbeck/Sanson CRISPRi/a-tailored rules, and select 5-6 guides per gene biased toward the window center. ```python def crispri_window(tss_coord, strand='+'): '''Dolcetto convention: search -50 to +300 around the FANTOM5 highest-rank CAGE peak. Reason: Sanson 2018 found +25 to +75 nt downstream of the TSS optimal for CRISPRi, so rank candidates toward that band; the search is relaxed outward to fill the per-gene guide quota when poorly-annotated TSSs leave too few candidates.''' if strand == '+': return (tss_coord - 50, tss_coord + 300) return (tss_coord - 300, tss_coord + 50) def crispra_window(tss_coord, strand='+'): '''Calabrese convention: -150 to -75 upstream of TSS. Reason: dCas9-VP64 (and SAM, SunTag) activate maximally when bound just upstream of Pol II loading. Horlbeck v2 CRISPRa uses -550 to -25 (broader, lower per-guide signal). For SAM, prefer Calabrese tightness; for SunTag, Horlbeck width is acceptable.''' if strand == '+': return (tss_coord - 150, tss_coord - 75) return (tss_coord + 75, tss_coord + 150) ``` **Critical nuance:** Cell-type-specific TSSs differ from the FANTOM5 consensus in ~15% of genes. For tissue-specific screens (e.g., neuron, hepatocyte), re-derive TSSs from a matched CAGE / GRO-seq / PRO-seq dataset before locking guide positions, or knockdown efficiency drops several-fold. The single most common cause of "weak" CRISPRi hits is mis-positioned guides against an alternative TSS. ## Genome-Wide Library Selection | Library | Year | Modality | Size (genes x guides) | sgRNA rules | Notable | |---------|------|----------|-----------------------|-------------|---------| | GeCKOv2 | 2014 | Cas9 KO | ~19k x 6 (~123k) | Exon position + off-target specificity (predates Rule Set 1) | Older; legacy datasets still use it | | Avana | 2016 | Cas9 KO | 110,257 as published; DepMap screens a ~4-guide subset (Meyers 2017: 70,086 after filtering, 17,670 genes) | Rule Set 1 | Still the Broad's primary Cas9 library; CERES->Chronos changed in 2021, not the library | | Brunello | 2016 | Cas9 KO | ~19k x 4 (~77k) | Rule Set 2 + CFD | Modern standard for new screens | | TKOv3 | 2017 | Cas9 KO | ~18k x 4 (~71k) | Hart on/off-target | Bagel/BAGEL2-optimized | | Humagne | 2020 | enAsCas12a | ~19.8k x 1 dual-guide construct (~20k per set) | enAsCas12a rules | Compact Cas12a sets C and D | | Horlbeck CRISPRi v2 | 2016 | dCas9-KRAB | ~18k x 5 (~104k) | Horlbeck CRISPRi rules | First-gen, still widely used | | Dolcetto | 2018 | dCas9-KRAB | ~19k x 3 per set (114,061 across Sets A+B) | Horlbeck + Rule Set 2 | Modern CRISPRi standard | | Horlbeck CRISPRa | 2016 | dCas9-VP64 | ~18k x 5 (~104k) | Horlbeck CRISPRa rules | Original CRISPRa | | Calabrese | 2018 | dCas9-VP64 | ~18.9k x 3 per set (113,238 across Sets A+B) | Tight TSS window | Modern CRISPRa standard | | Inzolia | 2024 | enAsCas12a | ~49k arrays: 19,687 genes (2 arrays each) plus ~4,435 paralog pairs | enAsCas12a rules | Paralog-pair multiplex; ~30% smaller than a typical Cas9 library | | in4mer | 2024 | Cas12a (4-guide) | Custom | enAsCas12a multiplex | Triple/quadruple KO per cassette | **dAUC trajectory (essentiality benchmark):** GeCKOv2 < Avana < Brunello/TKOv3 (Doench 2016 + Hart 2017). Moving from 4 to 6 sgRNAs/gene gives diminishing returns; the larger gain is moving from Rule Set 1 to Rule Set 2. **Cost-coverage tradeoff:** A 77k-guide Brunello at 500x cells/sgRNA needs 38.5M cells in pool, scalable. A 117k-guide Calabrese at 500x needs 59M cells -- often the deciding factor against CRISPRa for difficult-to-grow lines. ## PAM Variants and Alternative Cas Enzymes | Enzyme | PAM | Spacer length | Best for | |--------|-----|---------------|----------| | SpCas9 (WT) | NGG | 20 nt | Standard pooled screens; broadest library support | | eSpCas9, SpCas9-HF1 | NGG | 20 nt | Lower off-target rate; use for therapeutic-grade nomination | | SpCas9-NG | NG | 20 nt | Expanded targeting (~4x coverage); accept lower activity per guide | | SpRY | NRN / NYN | 20 nt | Near-PAMless; coverage at every position; ~50% lower per-guide activity | | SaCas9 | NNGRRT | 21 nt | AAV-packageable (small ORF); rarely used in pooled screens | | AsCas12a, LbCas12a | TTTV | 23 nt | AT-rich regions; staggered cut; lower expression noise | | enAsCas12a (DeWeirdt 2021) | Expanded TTTV + several non-canonical | 23 nt | Combinatorial / paralog screens | **Decision rule:** If the screen requires every possible TSS position (saturation tiling, dense regulatory dissection), use SpRY despite lower activity; otherwise, NGG is best because the on-target predictors were trained on it. ## Control Guides A genome-wide library should include: | Control type | Count | Purpose | |--------------|-------|---------| | Non-targeting (scrambled, no genomic match) | 500-1,000 (~1% of library) | Primary null distribution for CRISPRi/a; safe baseline for normalization | | Safe-harbor (AAVS1, ROSA26-equivalent) | 50-100 | Cas9-only: absorbs cut-toxicity baseline (matters for amplicon-correction) | | Olfactory receptors (presumed non-expressed) | 50-100 | Second null set for orthogonal normalization | | Reference essentials (CEGv2 subset: e.g. RPS3, RPL11, EIF3A, POLR2A) | 50-100 | Internal positive control; QC dropout signal | | Reference non-essentials (NEGv1 subset) | 50-100 | Internal negative control; BAGEL2 calibration | **Critical pitfall:** Using only AAVS1 as the negative control in a Cas9 screen creates a normalization baseline biased toward "any cut is bad." Always add NTCs or non-essentials so that downstream median normalization and PR-AUC against CEGv2 work without baseline-shift artifacts. ## Library Composition for Specialized Screens **Paralog buffering (Cas12a multiplex):** Build 4-guide arrays where positions 1-2 target gene A and positions 3-4 target paralog gene B. Inzolia covers ~4,435 paralog pairs within ~49k arrays. Singleton controls (gene A alone, gene B alone) must be included to score genetic interaction = double_KO_LFC - sum(single_KO_LFC). **Base editor screens (tiling-library design):** Tile NGG-adjacent spacers across exons; ensure editing window (positions 4-8 from PAM-distal end) lands inside coding exons; flag bystander Cs/As in the window for downstream interpretation. Restrict to 50-90% editing efficiency a priori (filter out predicted low-efficacy guides) -- see [[base-editing-analysis]]. **Tiling / regulatory dissection:** Dense (every 5-10 bp) CRISPRi or CRISPRa guides across the candidate region; CRISPRi has broader signal width (good for enhancer discovery) but Cas9-indel tiling has sharper resolution (good for pinpointing critical bases). Pair with CRISPR-SURF deconvolution. ## Oligo Design for Pooled Synthesis **Goal:** Generate the final oligo sequence ready for chip-based synthesis. Vendor limits differ: Twist oligo pools cap at ~300 nt per oligo with no fixed pool size, GenScript's 92K format spans 20-170 nt, and Agilent OLS 244K spans 30-230 nt. **Approach:** Add subpool PCR primers (so multiple sublibraries can share a synthesis array), the BsmBI/Esp3I overhang for golden-gate cloning into LentiGuide-Puro (Addgene 52963) or LentiCRISPRv2, and append the tracrRNA scaffold if the array length permits. ```python def build_oligo(spacer, vector='lentiGuide-Puro', subpool_idx=None): '''Construct final oligo for pooled synthesis. LentiGuide-Puro / LentiCRISPRv2 use BsmBI (Esp3I) with these overhangs: forward: 5'-CACCG[spacer]-3' reverse: 5'-AAAC[revcomp(spacer)]C-3' For chip synthesis, the spacer is flanked by subpool-specific PCR primers.''' subpool_fwd = { 1: 'GGAAAGGACGAAACACCG', # subpool 1 forward primer + BsmBI overhang 2: 'GAGGCACTGGGCAGGTACCG', }.get(subpool_idx, 'GGAAAGGACGAAACACCG') # First 33 nt of the Chen 2013 sgRNA(F+E) optimized scaffold. NOTE: lentiGuide-Puro (#52963) # and lentiCRISPRv2 (#52961) carry the ORIGINAL scaffold; F+E belongs to lentiCRISPRv2-Opti (#163126).