# Full existing prime-editing panel: sequence and product acceptance

SEQUENCE VERIFICATION, NOT EMPIRICAL EDITING EFFICIENCY.

The original item33 requirement is to verify existing pegRNA/nicking configurations and sequence provenance, with a candidate panel, computational rationale and unresolved empirical efficiency. **All6,480 existing rows now pass the expanded local sequence/product checks**, extending the earlier lead-only result. No candidate was added or recommended for therapeutic use.

## Exact coverage

`pe_panel_product_check.py` reads the retained `20_ANALYSIS/editing/prime_editing/PEG_RNA_DESIGNS.csv` and versioned NG_008934.1 FASTA. Source SHA-256 and all coverage counts are in `PE_PANEL_PRODUCT_RESULTS.json`. It covers30 distinct spacers,3,180 forward-strand and3,300 reverse-strand rows. Configuration counts are2,200 no-second-nick,2,200 PE3 and2,080 PE3b-labelled rows.

The6,480 unique IDs each pass:

- Stated pegRNA nick coordinate and nick-to-edit distance.
- Exact PBS annealing sequence, including the reverse-strand convention.
- RTT coverage of the target and copied product equal to the reference interval.
- Exactly one intended mutant-to-reference difference, at forward38,630,392.
- An independent position-by-position, strand-oriented RTT reconstruction in addition to contiguous interval extraction.
- Post-edit homology length, extension sequence/length, full canonical pegRNA concatenation and the stored full epegRNA concatenation.
- Consistency of the reference coding-strand CGG/Arg endpoint with the recorded output annotation.

For all4,280 rows with a second nick, the opposite-strand guide sequence and PAM match the repaired reference, and nick coordinate/distance agree. All2,080 PE3b-labelled rows overlap the edited base and have precisely one mutant-versus-repaired spacer mismatch. All2,200 no-nick rows have empty second-guide/PAM fields. There were zero failed rows. These checks do not independently establish the biological correctness of every scaffold or RNA-stabilizing element; concatenation agreement is the limited claim.

## Candidate panel and rationale

The accompanying JSON supplies all**88 existing spacer/system families**, with member count, representative existing ID, retained editor label, PBS/RTT lengths, nick-to-edit distance, post-edit homology, heuristic score and every tied top ID. Family counts sum exactly to6,480. This is a compact panel over the original table, not a new design collection.

The representative is selected using the table's existing `predicted_relative_efficiency` column solely to reconcile the old ranking. That field is a heuristic multiplied score, **not a percentage or calibrated efficiency**. Ties are retained rather than inventing a unique optimum. No cross-system biological superiority follows from the representative choice.

The computational rationale remains explicit and limited: PBS must complement the priming segment; RTT must encode the intended reference repair with stated homology; optional opposite-strand guides must match the intended post-repair local sequence. These requirements distinguish valid sequence bookkeeping from whether cells will actually produce that outcome. The manuscript's no-second-nick lead recommendation remains distinct from its first table row's PE3/PE5 configuration. This audit does not silently equate them.

## Independence, provenance and limits

This code does not invoke the original design generator or trust its product-success booleans. It reconstructs the copied sequence from the saved RTT and compares directly with local reference bases, then separately constructs the oriented sequence one coordinate at a time. The reference is anchored by a unique exact62-base window whose stated coordinates were already checked. This is not a fresh RefSeqGene-to-whole-hg38 alignment.

The earlier `EDITOR_CANDIDATE_AUDIT.md` supplied all-panel spacer/PAM identity and provenance checks, retained PRIDICT geometry context and specificity counting-unit reconciliation. This new result adds the missing all-panel product/PBS/opposite-nick scope. It does not rerun a trained efficiency predictor, verify its transport to cardiomyocytes, or recompute genome-wide specificity.

A single repaired/mutant mismatch in a PE3b-labelled spacer is **not proof of repair-only nicking**. Actual mismatch tolerance, editor/PAM behavior, per-position editing, indels, scaffold processing, delivery, safety and tissue efficacy remain empirical. The table's qualitative indel-risk text was not newly validated or endorsed. No clinical prediction follows from successful reference repair on paper.

The bounded item33 candidate/configuration deliverable is now complete at the requested computational sequence-audit level. No canonical source, candidate table or frozen output was modified; all new results remain in this program directory.