troubleshooting
Small RNA sequencing overview
Why small-RNA libraries fail when adapter dimers dominate, how miRNA arm bias and reference mapping differ from mRNA analysis, and how degradation can imitate a
- Author
- EVRINTH Editorial Team
- Published
- 8 October 2026
- Updated
- 8 October 2026
- Reading time
- 9 min

Adapter dimers are short, eager to form, and happy to occupy a sequencer. In a small-RNA library they are the first troubleshooting fact, because the insert you wanted is also short, and the two products compete. This page is the outline of small-RNA sequencing for a research bench that has a size trace it does not trust, or a count table that looks like an mRNA experiment and is not one. Mature mRNA library choices are a different selection, described in RNA-seq library types: poly(A) and ribodepletion. The path from cells to any expression result still starts at from cells to a gene expression result.
The decision in front of a size trace
You are deciding whether this library contains a small-RNA insert worth sequencing or analysing, and whether a later differential table describes microRNAs or a pile of fragments from broken longer RNA. The decision uses three observations: the input RNA's integrity, the library's size profile, and where the reads go when they are mapped. A large FASTQ file does not answer any of the three. Short-read chemistry as a class is sketched in the Illumina sequencing overview. The insert you selected is the part that makes the run a small-RNA run.
What "small" includes, and how the library grabs it
MicroRNAs are about twenty-two nucleotides and come from hairpin precursors. Cells also carry other short species, including piRNAs in some tissues, endogenous fragments with regulatory names, and a large background of pieces of transfer RNA, ribosomal RNA and messenger RNA. A small-RNA protocol does not know those names. It keeps molecules in a size window and ligates adapters to their ends so they can be amplified and sequenced.
Ligation is the biased step. Some end chemistries ligate more readily than others, so the reads over-represent sequences the adapters liked. That bias is protocol-specific. Two kits can rank the same biological sample differently. Compare samples only within one chemistry, and say which chemistry in the methods. Random bases in the adapter, or other bias-reduction designs, are attempts to flatten that preference. They are a method class, not a guarantee that every microRNA is seen in proportion to its molecules.
Size selection, by gel or by a bead scheme the protocol specifies, is what keeps the microRNA-sized insert and reduces both longer RNA and the adapter dimer. Exact cuts belong to the kit instructions. The decision-level point is that the window has to be checked on a trace before you commit a flow cell. An adapter dimer sits shorter than a microRNA-plus-adapter product. If the dimer peak is the tall one, the library is not ready.
Input RNA quality decides whether a small-RNA peak is plausible. Intact RNA can still be prepared into a small-RNA library that reflects the short species present in the cell. Badly degraded RNA donates fragments of longer transcripts into the same size window. Those fragments are real molecules and they are not the regulatory small RNAs you named. The extraction habits that keep longer RNA intact are in protecting RNA during extraction. They matter here because degradation manufactures false small RNA.
Mapping is not the mRNA pipeline with shorter reads
Aligners and counters built for long transcripts expect different error modes, different multi-mapping, and different annotations. MicroRNA analysis usually maps to a small-RNA reference of the miRBase class, or to the genome with a small-RNA annotation laid on top. Many mature microRNAs are nearly identical across families, so a read can fit several loci. A counter has to say whether those reads are discarded, assigned once, or split. That choice changes the table. IsomiRs, which are shifted or trimmed versions of a mature sequence, need a rule too. Collapsing them into the canonical name, or counting them separately, answers different questions.
Arm bias belongs in that rule. The 5-prime and 3-prime arms of a hairpin are different features. Biology often keeps one arm, and ligation bias can pretend to. Report the arm. A differential test between samples handled with one protocol can still see a biological change in arm use. A comparison across two ligation chemistries cannot, until the chemistry effect is measured.
Do not point an mRNA differential-expression script at a microRNA count matrix and accept the defaults for gene length, transcript models and filtering. Length normalisation designed for long genes is meaningless on a twenty-two nucleotide feature. Filters that drop "low counts" with an mRNA threshold can drop the entire microRNA set or keep only the few abundant ones. Use a contrast, a replicate structure and an error rate that you understand, with software settings written for small RNA. The differential expression analysis enquiry reference is a place to discuss a contrast in general. For small RNA, say explicitly that the features are microRNAs, or the conversation will assume messenger RNA.
Public sequence records in GenBank and archived reads in the European Nucleotide Archive show how variable small-RNA annotations and library descriptions are. Read the size-selection field before you reuse a count matrix.
Branches from a bad trace or a bad mapping rate
Start with the electropherogram or equivalent size profile. A dominant peak at the dimer length means stop. Go back to ligation conditions and size selection as the protocol allows, or rebuild from RNA that actually contains small-RNA mass. Sequencing that library spends the run on adapters. A peak at the expected insert length, with a minor dimer, can proceed, and the dimer reads should be recognised and set aside in the analysis rather than counted as genes.
If the trace looked right and most reads map to ribosomal RNA or to long-RNA fragments, the size window admitted degradation products or the input was not enriched for true small RNA. Check the integrity of the total RNA. If the sample was broken before the library, a new library from the same tube repeats the problem. Fresh material, handled cold, is the branch. If the RNA was intact and the reads still miss microRNA references, suspect ligation bias, the wrong species database, or a mapping setting that rejected multi-mappers you needed to keep under a stated rule.
If one sample's microRNA fraction collapses while its group-mates hold, treat it as a library failure first and a biological zero second. One empty column will dominate a heatmap. Remove or repeat that library before the contrast is interpreted.
| What you see | Likely source | Next check |
|---|---|---|
| Peak shorter than the microRNA product | Adapter dimer | Reselect size or rebuild; do not sequence a dimer-dominated library |
| Broad smear instead of a narrow insert | Degradation or missing size selection | Integrity of the input RNA, then a new preparation if it is broken |
| Reads map to ribosomal fragments | Longer RNA chopped into the window | Separate handling damage from a true small-RNA enrichment |
| One arm dominates every sample equally | Possible ligation preference | Compare only within one chemistry, and name the arm |
| Mature microRNAs absent, precursors suggested | Over-long inserts kept | Confirm the size window matches mature product |
| mRNA pipeline drops most features | Length and count filters from the wrong assay | Re-count with a small-RNA reference and a declared multi-map rule |
Failure modes that survive into a tidy table
A dimer that was not filtered can be annotated as an unknown feature and then "differentially expressed" because its ligation differed between batches. Filter on size and on adapter sequence before naming biology. A multi-mapping read assigned greedily to one family member will create a false specific change. Write the assignment rule. A sample sheet that uses mRNA gene symbols for microRNA arms will join the wrong annotation and look authoritative.
Cross-protocol meta-analysis is the subtle failure. Studies deposited years apart used different adapters and different size cuts. Their microRNA rankings disagree for technical reasons. A combined differential table without a chemistry covariate reports the kits. Within one study, biological replicates still set the n. Three libraries from one culture are one culture.
Research limits
Small-RNA sequencing here is a research measurement. It does not diagnose disease from a circulating microRNA list, and it does not set a biosafety level for the fluid or tissue you started with. Human specimens remain an ethics and institutional biosafety decision. Ligation reagents and gels have hazard notes. Follow the note for the kit class on your bench. A biomarker sentence needs a validation design this overview does not provide.
Transit that chopped the long RNA
A warm courier leg, a humid afternoon with tubes off ice, or a freezer that blinked during a power cut will fragment ribosomal and messenger RNA. Those pieces fall into the small-RNA size window and sequence beautifully. The trace looks "small" because the RNA was damaged, which is the opposite of a successful microRNA enrichment. Record the cold-chain handoff. If the integrity trace taken before shipping and the trace taken on arrival disagree, analyse the arrival tube as a damaged sample or replace it. Do not let a degradation smear become a microRNA signature in the differential table.
What to say when you ask for a small-RNA run
State the species, the tissue or biofluid, how the RNA was stabilised, whether you need a microRNA-sized window specifically, the biological replicates, and the contrast. Say that the analysis reference is a small-RNA catalogue and that adapter dimers are a stop condition. Reagent and instrument classes live in the genomics and sequencing catalogue. The sample path is the nucleic acid analysis pathway.
The mRNA sequencing enquiry reference is the neighbouring enquiry when the molecules you need are long and polyadenylated. The differential expression analysis enquiry reference can frame the contrast once the feature type is explicit. Put those facts on the quote request and ask whether a quotation is possible. A small-RNA method can be discussed from that note. The size trace remains the evidence that the discussion was about inserts rather than about dimers.
Questions from the bench
Why do adapter dimers take over a small-RNA library?
Small-RNA protocols ligate adapters to short inserts. When inserts are scarce, the adapters ligate to each other and that dimer is short, amplifies eagerly, and can dominate the size profile and the reads. Size selection and a dimer-aware cleanup are how the chemistry pushes dimers aside. A library that is mostly dimer is a failed library, even if the sequencer produced a large file.
Can I run a standard mRNA differential-expression pipeline on microRNA counts?
The statistical idea of a contrast and an error rate still applies, and the pipeline settings usually do not. MicroRNAs are tiny, multi-mapping, biased by ligation and by which arm of the hairpin was kept, and annotated in a reference of the miRBase class rather than as long transcripts. Copying an mRNA workflow without those changes mis-counts the molecules and mis-names the features.
How do I tell degradation products from genuine small RNA?
Genuine microRNAs occupy a narrow size window and map to known small-RNA hairpins or to a stated small-RNA reference. Degradation of longer RNA produces a smear of fragments, often from abundant ribosomal or messenger RNA, with a size profile that is not a clean microRNA peak. An integrity trace on the input RNA, plus the insert-size distribution of the library, separates those stories better than a differential table does.
What is arm bias?
A microRNA hairpin can yield a 5-prime arm and a 3-prime arm, and cells often load one arm into the silencing complex much more than the other. Ligation chemistry also prefers some ends over others, so the arm ratio in the reads mixes biology with protocol bias. Report which arm you counted, and do not treat a chemistry preference as a regulatory switch unless the bias was measured.
References
Manufacturer names identify published method classes. Trademarks remain with their owners. Catalogue records on this site are independent references for enquiry. They are not a statement of inventory, distribution rights or a supply commitment. This page is educational. It is not medical advice, a diagnostic protocol or a biosafety approval.
Catalogue
Related products and categories
These links follow the subject of the article into published manufacturer references. A listing is a reference for an enquiry, not a statement of stock or distribution rights.
Continue in this cluster
Related reading
From cells to a gene expression resultHow a laboratory goes from cells to a gene expression result, and how RNA quality decides between a focused RT-qPCR assay and RNA-seq.
RNA-seq library types polyA and ribodepletionHow poly(A) selection and ribosomal depletion keep different RNA molecules, and when degraded, bacterial or non-polyadenylated samples need the broader library.
Protecting RNA during extractionHow to keep RNA intact from lysis to elution: RNase control, what A260 ratios do not prove, and the checks that show whether the preparation is still usable.