What is being verified
The automated suite checks sequence orientation, coordinate conversion, exon ordering, amplicon extraction, input constraints, Primer3 invocation, paired binding and reference screening states. Controlled synthetic sequences are used to create unambiguous expected outcomes.
Property checks exercise invariants such as reverse complementation being reversible and returned primers matching their claimed template sites. Integration tests exercise the real locally installed Primer3 executable.
Evidence and current release status
Recorded on 2026-11-15: 93 automated Python tests passed, 8 browser scenarios passed against the production frontend and real backend, and 13 of 13 initial workflow benchmark cases passed. These are software and computational checks, not validated laboratory assays.
The recorded calculations used Primer3 2.6.1 and Ensembl release 116. Offline reference regressions cover human TP53, mouse Actb and zebrafish wnt10a, with both genomic and spliced templates and independently checked coordinates.
The browser scenarios cover actual candidate generation, gene lookup, copied sequences, CSV/FASTA/JSON exports, comparison, responsive and dark-mode views, keyboard access, invalid inputs and failed jobs. No independent Primer-BLAST comparison or wet-laboratory validation has been completed.
This public research preview has defined scientific limits. Broad mismatch-aware reference screening and independent scientific validation remain outstanding. Working public hosting and passing software tests do not establish experimental assay performance.
Reference-tool comparisons
Primer3 is the candidate-generation engine, so matching its recorded outputs verifies the adapter rather than independently validating the underlying algorithm. Primer-BLAST uses its own references, parameters and screening logic, so results can differ.
A useful benchmark records the exact tool versions, reference releases, inputs and mismatch policies. Differences should be investigated and reported, not hidden behind an aggregate pass percentage.
Known boundaries
Exact matching is not an exhaustive model of PCR binding. Ensembl annotations are a versioned reference, not a complete description of a particular biological sample. Computationally acceptable primers may still fail experimentally.
No claim of clinical approval, laboratory validation, universal specificity or predicted assay-success probability is made.