Guide · · By the tinysamples team
Sample naming conventions that survive real lab life.
A good sample ID is short, unique, permanent and meaningless: a prefix plus a counter, like S-0142, written the same way every time. Dates belong in a date column as YYYY-MM-DD, never inside the name — Excel has been rewriting gene names like SEPT2 into dates for years. This guide gives three formats to copy and the failure modes each avoids.
Why do sample names fail?
Sample names fail in three ways: they carry meaning that stops being true, they use characters people misread, and software rewrites them. All three are documented. The best-known case is Excel converting gene symbols like SEPT2 into dates — an error so common that the committee that names human genes renamed the genes.
The meaning problem first. A name like MS-KO-3-old-B encodes a strain, a treatment, a passage and an opinion. When any of those turns out to be wrong — the genotyping comes back different, the passage count was off by one — you either relabel the tube and orphan every earlier record, or you keep a name that lies. Meaning in the name is a liability, not a feature.
The character problem is smaller but constant: O and 0, I, l and 1, written in marker on the curved side of a frosted cryovial. Two people transcribe the same tube two ways, and now the freezer holds one sample with two histories.
The software problem is the famous one. A 2016 scan of 18 genomics journals found gene-name errors in roughly one fifth of papers with supplementary Excel gene lists — SEPT2 silently became 2-Sep, MARCH1 became 1-Mar. In 2020 the HUGO Gene Nomenclature Committee renamed 27 human genes to escape it: SEPT2 is now SEPTIN2, MARCH1 is MARCHF1. The genes changed because the software would not. Your sample names get no committee, so design them to survive Excel from day one.
What makes a good sample ID?
A good sample ID is short, unique, permanent and meaningless: a fixed prefix, a hyphen and a zero-padded counter, like S-0142. Because it says nothing about the sample, it can never become wrong. Everything with meaning — date, owner, contents, passage — lives in the record, keyed to the ID.
- Meaningless. Encode nothing. Not the project, not the date, not the condition. Facts change; the ID must not.
- Unique and permanent. One counter, assigned once, never reused — not even for a sample that was discarded on day two.
- Zero-padded. Unpadded numbers sort as text: S-100 files between S-10 and S-11. Pad to four digits and 9,999 samples sort in order.
- Unambiguous characters. Capital letters minus O and I, digits, one hyphen. No spaces, no slashes, nothing Excel can read as a date or scientific notation — identifiers like 2310009E13 come back as floating-point numbers.
- Dates in a date column, as YYYY-MM-DD. Never in the name. 03-04 is March or April depending on who reads it; 2026-03-04 is not.
Which ID format should you copy?
One of these three: a plain counter (S-0142) when the lab shares one list, a project prefix (TUM-0142) when several projects register samples independently, a year scope (2026-0142) when volume is high and age should be visible. Same rules in each case — the difference is who owns the counter.
| Fits when | Fails when | |
|---|---|---|
| Plain counter — S-0142 | One lab, one list, one counter. The default; start here. | Two people hand out numbers from memory, or two projects each start their own S- series. |
| Project prefix — TUM-0142 | Several projects register samples independently and each owns its counter. | Prefixes multiply, projects get renamed, or a sample moves projects and someone renames it — breaking permanence. |
| Year scope — 2026-0142 | High volume; the counter resets each January and the year is visible at a glance. | The year tempts people to read meaning back into the ID, and it looks enough like a date that the column must be formatted as text. |
Whichever you pick, the next free number must live in exactly one place. Our free sample-tracking spreadsheet template has the ID column, the zero-padding and the text formatting already set up.
What actually goes on the tube?
The ID, written large, and little else. A cryovial label offers about a square centimeter; the record has no limit. Contents, date, owner, concentration, location, every action since — all of that belongs in the record, keyed to the ID. Labels have to survive -80°C and ethanol. Records have to survive people.
The division of labor matters because each side fails differently. A meaning-stuffed label fails when the meaning changes; an ID-only label fails only when the ink does — so use cryogenic-rated labels or a solvent-resistant marker under clear tape, and test one tube overnight at -80°C with a 70% ethanol wipe before trusting a batch. And the record fails when it lives in one head: the tube labeled “good one” was perfectly clear to whoever wrote it. What the record itself needs — owner, location, status, a log — is its own guide.
How does tinysamples handle IDs?
We sell tinysamples, a lightweight sample tracker, and its answer to naming is to remove the decision: every sample gets an auto-generated sequential ID at registration, so nobody invents a scheme at the bench on a Friday afternoon. It is free for 100 samples, no card. If a disciplined spreadsheet covers you, the template above applies the same rules for nothing.