How it works & how to read the results ▼
Pipeline: your gene → its RefSeq mRNA (via mygene.info) → sequence + exon map from NCBI → Primer3 designs candidate pairs over the transcript. Nothing is pre-canned — primers are generated fresh for the current reference transcript.
Specific isoform? Instead of a gene name you can enter an exact transcript — a RefSeq accession (NM_000546) or an Ensembl id (ENST… human, ENSMUST… mouse) — and primers are designed on that transcript. Species is auto-detected for transcript IDs.
Exon-junction badge: EXON-JUNCTION = a primer sits directly on an exon–exon boundary, so it cannot bind the gene's own genomic DNA. SPANS INTRON = the amplicon crosses a junction, so genomic DNA would give a larger product. NO JUNCTION = both primers sit inside one exon, so the pair does not tell cDNA from genomic DNA.
How the junction is achieved: Primer3 is constrained onto the exon map, not just checked afterwards. On Prefer junction primer (the default) the transcript's junctions are first required to carry a primer; if none can, the amplicon is forced across a junction instead; only if that also fails do you get an unconstrained pair — and then the result says so explicitly rather than quietly handing back gDNA-unsafe primers. Require returns nothing rather than falling back, which is the right setting when a −RT control is not an option. Some genes have no junction to use at all: histones and interferons are single-exon, and there the honest answer is a −RT control, not a badge.
Processed pseudogenes — what the junction badge does not cover: a processed pseudogene is a retrotransposed copy of the mRNA. It has no introns, which means it contains the exon junctions, and a junction primer binds it perfectly. HNRNPU is the clean example: every pair Primer3 returns is exon-junction-spanning, and every one of them also amplifies HNRNPUP1 at exactly the same product size — one band, one melt peak, nothing to tell them apart.
These pseudogenes are usually partial copies: HNRNPUP1 is 842 bp against a 6,789 nt transcript and reproduces only positions ~2408–2907. So the fix is positional. Before designing, the copied window is located on the transcript and handed to Primer3 as an excluded region, which moves the primers off it — the same thing a commercially validated HNRNPU pair does by sitting at position 1862, outside the window. Pairs are then re-checked, and anything that still collides is demoted and flagged ⚠ PSEUDOGENE.
Sometimes both goals cannot hold at once — PTEN's exon junctions sit inside the very region PTENP1 copies. Avoiding the pseudogene wins that trade, and the result says why: a junction would not have protected against a processed pseudogene anyway, while genomic DNA is still caught by a −RT control. Genes with dozens of pseudogenes (GAPDH has 65) are handled against the first 10, and the result says so rather than implying a full sweep.
Why the five pairs sit apart: Primer3 ranks candidates on penalty alone, and penalty barely moves when a primer shifts by a few bases — so its top five were routinely one pair jittered (HNRNPU's first two differed by 3 nt). Pairs are now required to be at least 60 nt apart, which puts them on different exon junctions and means a pair that fails at the bench has genuine alternatives rather than near-copies that fail the same way. Where the transcript has no room for that — a short gene, or one boxed in by pseudogene footprints — the close pairs are still returned rather than handing you fewer options.
Defaults: Tm ≈ 60 °C (±2), length 18–25 nt, GC 40–60 %, GC-clamp, amplicon 70–200 bp — standard SYBR-Green qPCR settings.
Always verify: hit UCSC In-Silico PCR ↗ on any pair to confirm it amplifies a single, correct product against the genome before you order. Then grab the order CSV for IDT / Sigma bulk entry.