# Small RNA-Seq: Methods, Analysis, and Applications

## Introduction to Small RNA-Seq

Small RNA-seq is a high-throughput sequencing approach that profiles the full complement of small non-coding RNAs (sncRNAs) in a biological sample. Unlike standard mRNA-seq, which captures polyadenylated transcripts of hundreds to thousands of nucleotides, small RNA-seq is optimized for RNA molecules in the 18–40 nucleotide (nt) range. This size window encompasses several distinct classes of regulatory RNAs that share a common functional theme: sequence-specific silencing of complementary targets.

The purpose of small RNA-seq is twofold. First, it provides a quantitative snapshot of small RNA expression across conditions, enabling differential expression analysis. Second, it permits the discovery of previously unannotated small RNAs, including novel microRNAs (miRNAs) and PIWI-interacting RNAs (piRNAs). Because small RNAs are frequently dysregulated in disease and are stable in biofluids, small RNA-seq has become a cornerstone of [Biomarker Discovery](/knowledge/molecular-biology/biomarker-discovery) efforts and mechanistic studies of gene regulation.

### What Are Small RNAs?

Small RNAs are defined operationally by size and by their association with Argonaute (AGO) family proteins. The three principal classes are:

- **MicroRNAs (miRNAs):** ~22 nt RNAs that guide AGO proteins to partially complementary sites in messenger RNAs (mRNAs), typically in the 3′ untranslated region (UTR), leading to translational repression and mRNA destabilization.
- **PIWI-interacting RNAs (piRNAs):** 24–32 nt RNAs that associate with PIWI clade Argonaute proteins and silence [transposable elements](/knowledge/molecular-biology/transposable-element) in the germline.
- **Small interfering RNAs (siRNAs):** 21–23 nt RNAs with perfect complementarity to their targets, derived from double-stranded RNA (dsRNA) precursors, and central to RNA interference (RNAi).

Additional small RNA species include transfer RNA fragments (tRFs), small nucleolar RNA-derived RNAs (sdRNAs), and [Small Nuclear RNA](/knowledge/molecular-biology/small-nuclear-rna)-derived fragments. The distinction between these classes is not always absolute; for a more detailed comparison, see [Small RNA vs Microrna](/knowledge/molecular-biology/small-rna-vs-microrna).

### Why Sequence Small RNAs?

Small RNAs regulate gene expression at multiple levels—transcriptional, post-transcriptional, and translational—and are implicated in nearly every biological process, from development and differentiation to stress responses and oncogenesis. Sequencing small RNAs offers several advantages over hybridization-based methods such as microarrays: it is unbiased, can detect novel species, resolves closely related isoforms, and provides digital count data amenable to rigorous [statistical analysis](/blog/guides/statistical-analysis). Moreover, small RNA-seq can reveal processing intermediates and strand-specific information that microarrays cannot.

## Small RNA Biology and Regulatory Mechanisms

Understanding the biology of small RNAs is essential for interpreting sequencing data. Each class has a distinct biogenesis pathway, and the features of those pathways—such as 5′ phosphate status, 3′ end modification, and strand selection—directly influence library preparation strategies and bioinformatic annotation.

### miRNA Biogenesis and Function

miRNA genes are transcribed by RNA polymerase II as primary transcripts (pri-miRNAs) that contain a local hairpin structure. In the nucleus, the RNase III enzyme Drosha, in complex with the dsRNA-binding protein DGCR8, cleaves the pri-miRNA to release a ~70 nt precursor miRNA (pre-miRNA) with a 2-nt 3′ overhang. Exportin-5 transports the pre-miRNA to the cytoplasm, where Dicer, another RNase III enzyme, cleaves it near the loop to produce a ~22 nt duplex. The duplex is loaded into an AGO protein, and the passenger strand (miRNA*) is typically degraded, while the guide strand is retained.

The mature miRNA guides AGO to target mRNAs through partial base pairing. In mammals, seed region complementarity (nucleotides 2–8 of the miRNA) is the primary determinant of targeting. AGO-bound miRNAs recruit the GW182 family proteins, which in turn engage the CCR4-NOT deadenylase complex, promoting mRNA deadenylation, decapping, and degradation. The net effect is reduced protein output, with mRNA destabilization accounting for the majority of repression in most contexts.

### piRNA and Transposon Silencing

piRNAs are the most abundant small RNA class in the animal germline, where they defend genome integrity by silencing transposable elements. Unlike miRNAs, piRNAs are produced from single-stranded precursors in a Dicer-independent manner. Primary piRNAs are transcribed from genomic loci called piRNA clusters, processed by the endonuclease Zucchini (MitoPLD in mice), and loaded into PIWI proteins. The 3′ ends are trimmed and 2′-O-methylated by Hen1 (HENMT1 in mammals).

A secondary amplification loop, termed the ping-pong cycle, operates in the cytoplasm. A primary piRNA-guided PIWI protein cleaves a transposon transcript, generating a new piRNA with a 10-nt overlap at its 5′ end with the primary piRNA. This secondary piRNA is loaded into a different PIWI protein and can cleave complementary transcripts, amplifying the silencing signal. The ping-pong signature—a 10-nt overlap bias between sense and antisense piRNAs—is a key bioinformatic feature used to identify piRNA loci.

### siRNA and RNA Interference

siRNAs are derived from long dsRNA precursors. In plants and invertebrates, dsRNA is cleaved by Dicer into 21–23 nt siRNAs that load into AGO proteins. Because siRNAs have perfect complementarity to their targets, they direct AGO-mediated cleavage of the target mRNA—a process called RNA interference. In mammals, the RNAi pathway is largely restricted to the germline and to the response to exogenous dsRNA, although endogenous siRNAs have been described in mouse oocytes. For a deeper treatment, see [Small Interfering RNA](/knowledge/molecular-biology/small-interfering-rna).

## Library Preparation for Small RNA-Seq

Library preparation for small RNA-seq is technically demanding because the input molecules are short, have defined 5′ phosphates and 3′ hydroxyls, and lack the poly(A) tails that simplify mRNA-seq library construction. The standard protocol involves sequential ligation of adapters to both ends, followed by reverse transcription and PCR amplification.

### Size Selection and Purification

Total RNA is first size-fractionated to enrich for small RNAs. The most common approach is denaturing polyacrylamide gel electrophoresis (PAGE), where RNA is separated on a 15% urea gel and a gel slice corresponding to 18–40 nt is excised. Alternatively, commercial kits use silica membrane columns with ethanol-based binding buffers that preferentially retain small RNAs. PAGE offers superior size resolution and is the gold standard, but column-based methods are faster and more reproducible for high-throughput workflows.

A critical consideration is that the size selection window must be adjusted for the adapter ligation steps that follow. Because the final library includes adapter sequences (~40–60 nt total), the initial RNA size selection must account for the fact that only the insert portion will be sequenced. For example, if the sequencing read length is 50 nt and the adapters contribute 30 nt, the insert must be at least 20 nt to avoid sequencing through into the 3′ adapter.

### Adapter Ligation and RT-PCR

The core of small RNA library construction is the sequential ligation of a 3′ adapter and a 5′ adapter, both of which contain priming sites for reverse transcription and PCR. The 3′ adapter has a pre-adenylated 5′ end and a blocked 3′ end to prevent self-ligation. Ligation is catalyzed by T4 RNA ligase 2, truncated (T4 Rnl2tr), which ligates the pre-adenylated adapter to the 3′ hydroxyl of the small RNA but cannot ligate the adapter to itself. The 5′ adapter is ligated using T4 RNA ligase 1, which requires a 5′ phosphate on the RNA—a feature that is naturally present on mature miRNAs and piRNAs but absent from some degradation products.

After ligation, the product is reverse-transcribed using a primer complementary to the 3′ adapter, then PCR-amplified with primers that add the sequences required for cluster generation on the sequencing platform. The PCR cycle number should be kept low (typically 12–15 cycles) to minimize amplification bias and the formation of chimeric products.

### Avoiding Adapter Dimers and Bias

Adapter dimers—ligation products of the 3′ and 5′ adapters without an RNA insert—are the most common artifact in small RNA-seq libraries. They are shorter than legitimate products and can dominate the sequencing output if not removed. Strategies to minimize adapter dimers include:

1. **Use of pre-adenylated 3′ adapters** with a blocked 3′ end, which prevents adapter–adapter ligation.
2. **Size selection after ligation** to remove the ~40 nt adapter dimer band before PCR.
3. **Optimization of the RNA:adapter molar ratio**, typically 1:1 to 1:3, to favor productive ligation.
4. **Post-PCR size selection** using PAGE or AMPure XP beads to remove residual dimers.

Another source of bias is differential ligation efficiency: T4 RNA ligases have sequence preferences, particularly at the 5′ end of the RNA. This can skew the representation of certain small RNAs. Random nucleotide barcodes in the adapters can help identify and correct for such biases computationally.

## Sequencing Platforms and Experimental Design

The choice of sequencing platform and experimental design has a profound impact on data quality and interpretability. Small RNA-seq has specific requirements that differ from mRNA-seq.

### Read Length and Depth Requirements

For most small RNA classes, a read length of 50 base pairs (bp) single-end is sufficient. This allows the full insert to be read plus enough adapter sequence to confirm the insert boundaries. Longer reads (75–100 bp) are unnecessary for standard small RNA analysis but may be useful for detecting 3′ end modifications or for sequencing longer tRFs.

Sequencing depth requirements depend on the biological question. For differential expression of abundant miRNAs, 5–10 million reads per sample is generally adequate. However, detecting low-abundance small RNAs, piRNAs, or novel miRNAs requires deeper sequencing—20–50 million reads per sample. It is important to note that a substantial fraction of reads (often 20–40%) may map to rRNAs, tRNAs, or other abundant non-coding RNAs, reducing the effective depth for the small RNAs of interest.

### Choosing a Sequencing Platform

Illumina platforms (MiSeq, NextSeq, HiSeq, NovaSeq) dominate small RNA-seq due to their high throughput, low error rates, and short-read capabilities. The choice among them is primarily a matter of throughput and cost per sample. Ion Torrent platforms use semiconductor sequencing and are an alternative when rapid turnaround is needed, but they have higher insertion/deletion error rates in homopolymer regions, which can complicate alignment of short reads.

For most applications, Illumina single-end 50 bp sequencing is the recommended default. Paired-end sequencing is not necessary for small RNAs because the inserts are shorter than the read length; paired-end reads would simply overlap.

### Experimental Controls and Replicates

Biological replicates are essential for reliable differential expression analysis. A minimum of three biological replicates per condition is recommended, with more replicates providing greater [statistical power](/blog/guides/statistical-power-what-it-is-and-why-it-matters-in-research). Technical replicates (sequencing the same library twice) are less valuable and are generally unnecessary given the low technical variability of [Illumina sequencing](/knowledge/diagnostics/molecular/illumina-sequencing-principle-chemistry-and-workflow).

Important controls include:

- **Input RNA controls:** Synthetic spike-in RNAs (e.g., from the External RNA Controls Consortium, ERCC) added at known concentrations to each sample before library preparation. These allow assessment of technical variability and normalization.
- **No-template controls:** A library prepared without RNA to detect adapter dimer contamination.
- **[RNA integrity assessment](/knowledge/diagnostics/molecular/rna-integrity-assessment-rin-values-gel-electrophoresis):** Bioanalyzer or TapeStation traces to confirm the presence of small RNA peaks and the absence of degradation.

## Bioinformatics Analysis of Small RNA-Seq Data

The computational analysis of small RNA-seq data follows a pipeline that shares steps with mRNA-seq but has critical differences, particularly in alignment and annotation.

### Preprocessing and QC

Raw sequencing reads must first be assessed for quality using tools such as FastQC. Key metrics include per-base quality scores, GC content, and the presence of adapter contamination. Because small RNA inserts are shorter than the read length, adapter sequences will appear at the 3′ end of many reads; these must be trimmed before alignment.

Adapter trimming is performed with tools such as Cutadapt or Trimmomatic. The trimming command should specify the adapter sequence, a minimum read length (typically 18 nt after trimming), and a maximum error rate (usually 0.1). After trimming, reads are collapsed to unique sequences with counts, which reduces computational burden and simplifies downstream analysis.

### Alignment to Reference Genome

Small RNA reads are short, so alignment requires a strategy that tolerates multi-mapping reads. Many small RNAs, particularly piRNAs and tRFs, map to multiple genomic locations. The choice of aligner matters: Bowtie and Bowtie2 are commonly used, with Bowtie1 (allowing zero mismatches) being preferred for miRNA analysis because it handles short reads efficiently. STAR is less commonly used for small RNAs because it is optimized for spliced alignment of longer reads.

A recommended strategy is to align reads first to a reference genome with no mismatches allowed. Reads that fail to map can then be aligned with one mismatch. Multi-mapping reads should be handled explicitly: either discard them, distribute their counts proportionally, or retain them with a note that their genomic origin is ambiguous.

### Quantification and Differential Expression

Quantification of known small RNAs is performed by counting reads that overlap annotated features. For miRNAs, the primary annotation source is miRBase, which provides mature miRNA sequences and genomic coordinates. For piRNAs, piRBase and RepeatMasker annotations are used.

Differential expression analysis uses count-based statistical models. The most widely used tools are DESeq2 and edgeR, both of which model count data with negative binomial distributions. These tools accept a count matrix (genes × samples) and a design matrix describing the experimental conditions. For small RNA-seq, it is important to filter out features with very low counts before testing, as these contribute noise and increase the multiple-testing burden.

## Annotation and Identification of Novel Small RNAs

A major advantage of small RNA-seq is the ability to discover novel small RNAs. This requires careful annotation of known features and the application of structural and sequence criteria to predict new ones.

### Known Small RNA Annotation

The first step is to annotate reads that map to known small RNA loci. For miRNAs, this involves checking whether the read corresponds to the mature or star strand and whether the 5′ end matches the annotated position. For piRNAs, reads are annotated by overlap with piRNA clusters or RepeatMasker-annotated transposable elements. Reads mapping to rRNAs, tRNAs, snoRNAs, and snRNAs are typically classified as "other" and may be excluded from downstream analysis unless the study focuses on tRFs or sdRNAs.

### Predicting Novel miRNAs

Novel miRNA prediction relies on the characteristic features of miRNA biogenesis. The key criteria are:

1. The candidate locus must form a stable hairpin structure (minimum free energy ≤ −25 kcal/mol).
2. The mature miRNA sequence must be located in one arm of the hairpin.
3. The hairpin must have a 2-nt 3′ overhang, consistent with Drosha and Dicer processing.
4. The mature miRNA should be present in the sequencing data at high abundance relative to the star strand.

Tools such as miRDeep2 and miRDeep-P2 implement these criteria. They score candidate hairpins based on the distribution of reads along the precursor, the thermodynamic stability of the hairpin, and the presence of the 2-nt overhang. Novel miRNAs should be validated by Northern blotting or RT-qPCR before being reported.

### piRNA Clusters and Ping-Pong Signatures

piRNA identification is more complex because piRNAs do not have a conserved precursor structure. Instead, piRNAs are identified by their genomic organization into clusters and by the ping-pong signature. The ping-pong signature is detected by examining the overlap between sense and antisense piRNAs: in a canonical ping-pong cycle, the 5′ ends of sense and antisense piRNAs overlap by exactly 10 nt. Tools such as proTRAC and piPipes can identify piRNA clusters and assess ping-pong enrichment.

## Data Normalization and Differential Expression

Normalization is critical for comparing small RNA expression across samples, but small RNA-seq data present unique challenges that make standard mRNA-seq normalization approaches problematic.

### Normalization Strategies

The simplest normalization is counts per million (CPM), which divides each count by the total number of mapped reads and multiplies by one million. CPM is easy to compute but assumes that the total RNA content is constant across samples—an assumption that is often violated in small RNA-seq because a few highly abundant miRNAs can dominate the library.

Transcripts per million (TPM) is a length-normalized metric that accounts for transcript length. However, length normalization is irrelevant for small RNAs, which are all approximately the same length, and TPM does not address the composition bias problem.

For differential expression analysis, DESeq2 and edgeR use median-of-ratios and trimmed mean of M-values (TMM) normalization, respectively. These methods estimate size factors that account for differences in library composition. They are generally preferable to CPM or TPM for small RNA-seq data, particularly when the composition of the small RNA population differs substantially between conditions.

### Differential Expression Testing

DESeq2 and edgeR both model count data with negative binomial distributions and provide shrinkage estimates of dispersion. For small RNA-seq, it is important to note that the dispersion estimates can be noisy when the number of replicates is small. Both tools provide methods for sharing information across features (empirical Bayes shrinkage) that improve dispersion estimation.

The output of differential expression analysis is a list of features with log2 fold changes, p-values, and adjusted p-values (controlling the false discovery rate). A common threshold for significance is adjusted p < 0.05 and |log2 fold change| > 1.

### Handling Low Counts and Zeros

Small RNA-seq data contain many features with zero or very low counts, particularly for piRNAs and rare miRNAs. These zeros are problematic for several reasons. First, they inflate the variance and reduce [statistical power](/blog/guides/statistical-power-what-it-is-and-why-it-matters-in-research). Second, they can arise from technical dropout (failure to capture a small RNA that is present) or from genuine absence. Most differential expression tools handle zeros by adding a small pseudocount (e.g., 1) or by modeling the zero-inflation explicitly. For exploratory analysis, it is often useful to filter features with low mean counts (e.g., < 10 counts across all samples) before testing.

## Common Pitfalls and Best Practices

Small RNA-seq is prone to specific technical and analytical failures. Recognizing these pitfalls is essential for producing reliable data.

### Technical Artifacts

- **Adapter dimers:** The most common artifact, adapter dimers consume sequencing reads and skew quantification. They are identified during QC as a prominent peak at ~40 bp in the size distribution. Prevention includes careful size selection and optimization of ligation conditions.
- **Ligation bias:** T4 RNA ligases have sequence preferences that can distort the representation of small RNAs. This is particularly problematic for miRNAs with certain 5′ terminal nucleotides. Random barcodes in adapters can help identify and correct for this bias.
- **RNA degradation:** Small RNAs are relatively stable, but rRNA degradation fragments can contaminate the small RNA fraction. Bioanalyzer traces should be inspected for the presence of degradation products.
- **Index hopping:** On patterned flow cells (e.g., NovaSeq), index sequences can be misassigned between samples. Using unique dual indexes mitigates this risk.

### Analysis Mistakes

- **Incorrect adapter trimming:** Failing to trim adapters or trimming too aggressively can lead to misalignment and loss of reads. The trimming parameters should be validated on a subset of reads.
- **Ignoring multi-mapping reads:** Many small RNAs map to multiple loci. Discarding all multi-mappers loses information, while retaining all of them can inflate counts. A balanced approach is to report both uniquely mapping and multi-mapping counts.
- **Using mRNA-seq normalization:** Applying TPM or CPM normalization without considering composition bias can lead to false positives and negatives in differential expression analysis.
- **Overlooking batch effects:** Batch effects—systematic technical variation between sample batches—can confound biological differences. For a detailed discussion, see [Combat Batch Effect Removal](/knowledge/molecular-biology/combat-batch-effect-removal).

### Best Practices for Reproducibility

1. **Document all parameters:** Record the exact versions of all software and the parameters used for trimming, alignment, and quantification.
2. **Use a containerized pipeline:** Tools such as Docker or Singularity ensure that the analysis environment is reproducible.
3. **Include spike-in controls:** Synthetic RNAs added at known concentrations allow assessment of technical variability and can serve as a normalization reference.
4. **Validate key findings:** Confirm differential expression of key small RNAs by RT-qPCR or Northern blotting.
5. **Deposit data and code:** Make raw sequencing data available in public repositories (e.g., GEO, SRA) and share analysis code.

| Step | Common Error | Consequence | Mitigation |
|------|--------------|-------------|------------|
| Size selection | Window too broad | rRNA contamination | Tighten gel slice to 18–40 nt |
| Adapter ligation | Adapter dimer formation | Wasted reads | Pre-adenylated adapters, size selection |
| PCR amplification | Excessive cycles | Amplification bias | Limit to 12–15 cycles |
| Trimming | Adapter not removed | Misalignment | Validate trimming on subset |
| Alignment | Multi-mappers discarded | Loss of piRNA signal | Report unique and multi-mapping counts |
| Normalization | CPM without filtering | Composition bias | Use DESeq2/edgeR size factors |

## Frequently Asked Questions

### What is the difference between small RNA-seq and mRNA-seq?

Small RNA-seq targets RNA molecules of 18–40 nt, including miRNAs, piRNAs, and siRNAs, while mRNA-seq targets polyadenylated messenger RNAs of hundreds to thousands of nucleotides. The library preparation differs fundamentally: small RNA-seq requires adapter ligation to both ends of the RNA, whereas mRNA-seq typically uses poly(A) selection or rRNA depletion followed by random priming. The bioinformatic analysis also differs, with small RNA-seq requiring adapter trimming and short-read alignment, while mRNA-seq involves splice-aware alignment and transcript assembly.

### How do I choose the right sequencing depth for small RNA-seq?

The required depth depends on the abundance of the small RNAs of interest. For profiling abundant miRNAs, 5–10 million reads per sample is usually sufficient. For detecting low-abundance miRNAs, piRNAs, or novel small RNAs, 20–50 million reads per sample is recommended. A useful rule of thumb is to sequence enough to achieve saturation: the number of new small RNA species detected should plateau as sequencing depth increases.

### What are adapter dimers and how do I avoid them?

Adapter dimers are ligation products of the 3′ and 5′ adapters without an RNA insert. They are shorter than legitimate library fragments and can dominate sequencing output. They are avoided by using pre-adenylated 3′ adapters with blocked 3′ ends, optimizing the RNA:adapter ratio, performing size selection after ligation, and using post-PCR size selection to remove residual dimers.

### Which bioinformatics tools are best for small RNA-seq analysis?

There is no single best tool; the choice depends on the analysis stage. For QC, FastQC and MultiQC are standard. For adapter trimming, Cutadapt is the most widely used. For alignment, Bowtie1 (with zero mismatches) is preferred for miRNA analysis, while Bowtie2 is more flexible. For quantification and differential expression, DESeq2 and edgeR are the most robust. For novel miRNA prediction, miRDeep2 is the standard tool.

### How do I normalize small RNA-seq data?

For differential expression analysis, use the size-factor normalization implemented in DESeq2 or edgeR, which accounts for differences in library composition. CPM or TPM are acceptable for exploratory analysis but can be biased when the composition of the small RNA population differs between samples. Spike-in RNAs can be used as an alternative normalization reference, but they require careful calibration.

### Can small RNA-seq detect piRNAs?

Yes, small RNA-seq is the primary method for detecting piRNAs. However, piRNAs are 24–32 nt, so the size selection window must include this range. Additionally, piRNAs are often derived from transposable elements and map to multiple genomic locations, so the alignment strategy must handle multi-mapping reads. The ping-pong signature can be used to confirm that detected piRNAs are functional.

### What are common mistakes in small RNA-seq analysis?

Common mistakes include: failing to trim adapters before alignment, discarding multi-mapping reads (which loses piRNA information), using mRNA-seq normalization methods without considering composition bias, ignoring batch effects, and failing to validate novel small RNAs experimentally. Technical mistakes include adapter dimer contamination, excessive PCR amplification, and RNA degradation.

## Key Takeaways

- Small RNA-seq profiles 18–40 nt regulatory RNAs, including miRNAs, piRNAs, and siRNAs, and requires specialized library preparation with sequential adapter ligation.
- The biogenesis pathways of each small RNA class—Drosha/Dicer for miRNAs, Zucchini and the ping-pong cycle for piRNAs, and Dicer for siRNAs—dictate the features used for bioinformatic annotation.
- Library preparation is the most error-prone step, with adapter dimer formation and ligation bias being the primary technical challenges.
- Sequencing depth should be matched to the biological question: 5–10 million reads for abundant miRNAs, 20–50 million for rare species and novel discovery.
- Bioinformatics analysis requires adapter trimming, short-read alignment with multi-mapper handling, and count-based differential expression using DESeq2 or edgeR.
- Normalization must account for library composition bias; size-factor methods are preferred over simple CPM or TPM.
- Novel miRNA and piRNA prediction requires structural criteria (hairpin stability, 2-nt overhang) and the ping-pong signature, respectively, and should be validated experimentally.

## Further Reading

- Benesova S, Kubista M, Valihrach L. *Small RNA-Sequencing: Approaches and Considerations for miRNA Analysis*. Diagnostics (Basel, Switzerland). 2021. [PubMed 34071824](https://doi.org/10.3390/diagnostics11060964)
- Raabe CA et al. *Biases in small RNA deep sequencing data*. [Nucleic acids research](/blog/news/nucleic-acids-research). 2014. [PubMed 24198247](https://doi.org/10.1093/nar/gkt1021)
- Kyriakidis I, Kyriakidis K, Tsezou A. *MicroRNAs and the Diagnosis of Childhood Acute Lymphoblastic Leukemia: Systematic Review, Meta-Analysis and Re-Analysis with Novel Small RNA-Seq Tools*. Cancers. 2022. [PubMed 36010971](https://doi.org/10.3390/cancers14163976)
- Coruh C, Shahid S, Axtell MJ. *Seeing the forest for the trees: annotating small RNA producing genes in plants*. Current opinion in plant biology. 2014. [PubMed 24632306](https://doi.org/10.1016/j.pbi.2014.02.008)
- Li J et al. *COMPSRA: a COMprehensive Platform for Small RNA-Seq data Analysis*. Scientific reports. 2020. [PubMed 32165660](https://doi.org/10.1038/s41598-020-61495-0)
- Kuksa PP et al. *SPAR: small RNA-seq portal for analysis of sequencing experiments*. [Nucleic acids research](/blog/news/nucleic-acids-research). 2018. [PubMed 29733404](https://doi.org/10.1093/nar/gky330)

## Related Clinical & Scientific Guides

* [MAPK Pathway: Mechanism, Function, and Clinical Relevance](/knowledge/molecular-biology/mapk-pathway)
* [Mammalian Cell Culture Bioreactors: A Practical Guide](/knowledge/molecular-biology/mammalian-cell-culture-bioreactor)
* [Nucleotide Formation: Biosynthesis and Assembly of DNA/RNA Building Blocks](/knowledge/molecular-biology/nucleotide-formation)