Primer design and a specificity check

A primer pair is a hypothesis about two binding sites and one amplicon. Pretty melting temperatures are not enough. You still have to ask: will these oligos bind where you think, and almost nowhere else, on the template you actually have?

Numbers you must be able to explain

Property	Typical classroom range	Why it matters
Length	18–25 nt	Shorter oligos lose specificity; longer ones can fold and cost more.
GC	~40–60%	Extreme GC shifts Tm and can create sticky oligos.
Tm	~55–65 °C, pair within ~2–5 °C	The pair has to anneal in the same cycle. Formulae differ; name the one you used.
Product size	assay-dependent	qPCR, cloning, and genotyping want different lengths.
3′ end	avoid long GC clamps / runs	Extension starts here; a sloppy 3′ end is a common off-target cause.

Tm is model-dependent (salt, nearest-neighbour vs GC%). If two tools disagree by a few degrees, write both and pick one consistently.

A design loop you can finish in one sitting

Choose the template: a CDS, an exon, or a short FASTA you can defend.
Note length, GC, and a rough Tm of the whole template so you know what “normal” looks like.
Generate candidate pairs under constraints (Tm, product size, primer length).
Reject pairs with long homopolymers, extreme GC, or a 3′ that looks like a clamp.
Only then run an in silico specificity check against the intended genome or a small panel.

# Conceptual check (any IUPAC-aware scanner)
# Forward 5'→3'   Reverse 5'→3'
# Ask: how many amplicons within the size window
#      with ≤ N mismatches per oligo?

Where software fits — after the biology

Primer3 remains the usual desktop/server starting point. NCBI Primer-BLAST is the common web check against a public database. For a first pass on a short sequence, Japality Primer can compute length, GC, reverse complement, and ranked pairs. That step is still only design. Specificity is a different question: an in silico PCR / scan such as japality-primer or a BLAST search against the intended genome asks where else the pair can bind. A pair that looks clean in a designer and dirty in a genome scan is not ready for an order sheet.

Failure modes that look like “bad luck”

The pair is fine; the template was the wrong transcript or the wrong assembly.
The pair is fine on paper; a SNP sits under the 3′ end in your samples.
The pair amplifies two paralogues. A Venn of “intended gene vs family” would have warned you.
You designed on cDNA coordinates and then ordered oligos for genomic DNA (or the reverse).

What to write before anyone pipettes

Template accession and sequence version
Forward / reverse (5′→3′), length, GC, Tm method
Expected product size and cycling sketch
Specificity method and the database or FASTA you used
A sentence on what would make you redesign

Next: formats if FASTA still feels fuzzy, or visualisation if you need to compare several candidate sets.