In-Silico Evaluation of CRISPR/Cas9 Guide RNAs Targeting Rice OsSWEET14 for Bacterial Blight Resistance

In-Silico Evaluation of CRISPR/Cas9 Guide RNAs Targeting Rice OsSWEET14 is an M.Sc biotechnology dissertation that designs, scores and shortlists genome-editing guides for a real agronomic target — the rice sucrose-effluxer gene whose promoter Xanthomonas oryzae pv. oryzae hijacks during bacterial blight, one of the most damaging diseases of rice in Tamil Nadu and across Asia — without touching a pipette.

The biology is well established: pathogenic Xanthomonas strains deliver TAL effectors that induce OsSWEET14 (also annotated Os11N3) expression, and disrupting the promoter's effector-binding elements or the gene itself confers resistance, as shown first with TALENs and later pursued with CRISPR/Cas9. What follows from that literature is a concrete computational question: which Cas9 guide RNAs in and around OsSWEET14 are most likely to edit efficiently, with minimal off-target risk, in the japonica reference genome?

This study answers it with a declared protocol rather than a casual tool session. The OsSWEET14 locus and promoter are retrieved from RAP-DB; target regions are defined in writing (promoter effector-binding region and coding exons); candidate NGG-PAM guides are generated with CRISPOR and CHOPCHOP; every candidate is scored for on-target efficiency (Doench 2016-style scores), specificity (CFD and MIT off-target scores against the reference genome), guide quality (GC content, homopolymers, poly-T runs, self-complementarity) and independent off-target verification by BLAST. The two tools' rankings are compared for concordance, and the study delivers a ranked shortlist with validation-amplicon designs — a lab-ready starting document for any department that later pursues wet-lab work, which is kept strictly outside this dissertation's scope.

Genome-editing experiments on rice fail or mislead at the design stage more often than at the bench stage: a guide chosen casually can edit inefficiently, or worse, cut at unintended sites that no one checks for. Guide-selection web tools make it easy to press a button, but easy tools do not make a defensible choice — target regions must be biologically justified, off-target policy must be declared before scores are seen, tool predictions must be verified independently, and the whole selection must be documented so that a colleague can reproduce it exactly.

For the bacterial-blight susceptibility gene OsSWEET14 specifically, no openly documented, protocol-driven guide evaluation exists for a student to study and extend; the published editing papers report conclusions only. There is a need for a reproducible in-silico evaluation that defines the target space from the TAL-effector literature, enumerates candidate SpCas9 guides with two independent tools, scores on-target potential and off-target risk under pre-declared thresholds, verifies predictions against the reference genome, quantifies agreement between tools, and hands forward a ranked, validation-ready shortlist with an honest account of what remains unverifiable by computation alone.

Abstract

Bacterial blight of rice, caused by Xanthomonas oryzae pv. oryzae, is a recurring constraint on paddy yields in Tamil Nadu. The pathogen's TAL effectors activate the host sucrose-transporter gene OsSWEET14 to feed the infection, making that locus a rational genome-editing target for durable resistance. This dissertation performs a complete, protocol-driven in-silico evaluation of CRISPR/Cas9 guide RNAs for OsSWEET14 in the japonica reference genome, producing a ranked, tool-cross-validated shortlist and an honest statement of what computational screening can and cannot guarantee.

Introduction

Genome editing has moved from journal pages to breeding programmes, and rice is its leading crop. But editing experiments are expensive: tissue culture, transformation and regeneration consume months, so the guide selection that precedes the bench work determines whether those months are spent well. In-silico guide evaluation is the standard first phase of any editing project, and exactly the kind of rigorous, examinable work an M.Sc dissertation can complete with free public tools and data.

Existing resources and the gap

Guide-design web tools are abundant — CRISPOR, CHOPCHOP and others — but tool output alone is not a study. Typical student exercises run one tool and quote its top hit, with no declared target definition, no off-target policy, no cross-tool comparison and no documented reproducibility. Published rice editing papers report only their final guides, not the screening that produced them. The gap this project fills: a transparent, documented evaluation protocol for OsSWEET14 — target regions justified from the TAL-effector literature, pre-declared scoring rules, two tools compared for concordance, BLAST verification of every off-target call, and versioned outputs that re-derive every table.

Proposed study

• Target definition: the OsSWEET14 locus retrieved from RAP-DB with coordinates recorded; target space fixed as (a) the promoter region carrying the TAL-effector binding context and (b) early coding exons, with the biological rationale of each written in the protocol.
• Guide generation: all SpCas9 NGG-PAM guides in the target space enumerated by both CRISPOR and CHOPCHOP.
• Scoring: on-target efficiency scores, CFD and MIT specificity scores, guide-quality flags (GC 40–60% window, homopolymers, poly-T terminators, self-complementarity), each threshold fixed before results are seen.
• Verification: every predicted off-target with up to three mismatches re-checked by independent BLAST against the rice genome.
• Concordance and shortlisting: agreement analysis between the two tools' rankings; a shortlist of guides that pass all thresholds in both pipelines; validation-amplicon design around each shortlisted guide.
• Deliverable: a screening report, ranked shortlist and reproducibility pack — a lab-ready starting document for a future wet-lab phase, which remains explicitly out of scope.

Feasibility

• Technical: all tools are free web services or open-source software; the rice reference genome is public; a laptop suffices.
• Economic: zero reagent cost — no culture media, enzymes or consumables.
• Operational: the analysis is fully schedulable in 20 weeks with no weather, growth or contamination risk; the main risk (tool service downtime) is mitigated by archiving all queries and outputs as files, which the reproducibility policy requires anyway.

In scope

• Computational evaluation only: guide enumeration, scoring, off-target prediction and verification for SpCas9 (NGG PAM) against the Oryza sativa ssp. japonica reference genome.
• The OsSWEET14 locus: promoter region with its TAL-effector binding context and early coding exons.
• Two design tools (CRISPOR, CHOPCHOP) compared for concordance, with BLAST as the independent verifier.
• A ranked shortlist with validation-amplicon designs, a screening report and a versioned reproducibility pack.

Out of scope

• All wet-lab work: Agrobacterium-mediated transformation, rice tissue culture, Cas9 construct assembly, regeneration and disease screening — these require biosafety approvals and facilities beyond the dissertation and are framed as the follow-on phase.
• Editing of other cultivars' genomes (indica pangenomes) beyond a note on cultivar-sequence divergence.
• Delivery-system design, promoters, selection cassettes and regulatory filing for edited plants.
• Other nucleases (Cas12a, base editors) beyond a comparative discussion paragraph.
• Field-level resistance prediction — that requires the wet-lab phase and glasshouse trials that follow it.

Research design: a protocol-driven computational screening study with a declared decision pipeline — no experiments, every step scripted or archived so the entire evaluation re-runs from the archived inputs.

Materials and tools: RAP-DB (reference sequences and annotation), CRISPOR (on-target and CFD/MIT off-target scoring), CHOPCHOP (independent enumeration with plant settings), NCBI BLAST (verification), Python 3 with Biopython for sequence handling and table assembly, Excel/R for the screening matrix. Analysis plan (pre-declared): GC window 40–60%; off-target consideration limited to guides with up to three mismatches; any off-target with two or fewer mismatches adjacent to a PAM within coding regions disqualifies a guide outright; shortlisting requires concordant passage of all thresholds in both tools. All thresholds carry their rationale in the protocol, and any deviation made during the work is logged with its reason rather than silently applied.

The evaluation is a five-layer pipeline, each layer versioned on disk: sequence archive, enumeration outputs, the screening matrix, verification tables and the final shortlist. Data flows one way, and every downstream artefact is derivable from the layer above it.

flowchart TD
  A["RAP-DB: OsSWEET14 locus, promoter, transcripts (FASTA + coordinates archived)"] --> B["Written protocol: target space, thresholds, off-target policy (approved first)"]
  B --> C["CRISPOR: guide enumeration + Doench/CFD/MIT scores"]
  B --> D["CHOPCHOP: independent guide enumeration (plant settings)"]
  C --> E["Screening matrix: efficiency, specificity, GC window, quality flags"]
  D --> E
  E --> F{"Guide passes all pre-declared thresholds?"}
  F -->|fails, reason recorded| G["Rejected-guides log"]
  F -->|passes| H["NCBI BLAST verification of every off-target call"]
  H --> I["Reconciliation: tool predictions vs BLAST"]
  I --> J["Concordance analysis of the two tools' rankings"]
  J --> K["Ranked shortlist + validation amplicons"]
  K --> L["Reproducibility pack: scripts, archives, run logs"]

Three design decisions carry the rigour. First, the protocol precedes the computation: the target space and every threshold are fixed in an approved document before any guide is enumerated, which is what stops the shortlist from being reverse-engineered from a favourite candidate. Second, two independent enumeration pipelines run in parallel; their concordance is itself a result — a guide that both tools rank well, for different internal reasons, is more trustworthy than either tool's top hit alone. Third, BLAST verification is independent of both tools' databases and algorithms, so a tool's missed off-target has a second chance of being caught; disagreements between tool and BLAST are resolved conservatively (the stricter call wins) and the resolution rule is written down. The rejected-guides log is kept deliberately — examiners ask what was excluded and why, and the log answers with evidence instead of memory.

• Execute the wet-lab phase under institutional biosafety approval: construct assembly, rice transformation and bacterial-blight screening of the shortlisted guides.
• Extend the evaluation to indica cultivar sequences to quantify cultivar-dependent guide performance before any varietal programme.
• Add a third enumeration tool and formally evaluate whether concordance across three pipelines improves shortlist quality.
• Compare SpCas9 with Cas12a (TTTN PAMs) for the same target space, widening the editable window.
• Model the TAL-effector binding context quantitatively, scoring promoter guides by predicted effector-site disruption rather than proximity alone.