mRNA Sequencing · Transcriptome Profiling

Profile the transcriptome. Discover expression changes.

End-to-end mRNA Sequencing and data analysis for plants, animals and microbes — comprehensive, quantitative profiling of the polyadenylated transcriptome for gene expression, differential expression and transcript-level characterisation.

Plan Your Project → See the Workflow
>95%
Mean alignment rate (demo)
>28K
Genes quantified (rice demo)
2
Modes: reference & de novo
DESeq2
+ edgeR for DE analysis
About the Service

Transcriptome profiling for any organism

Nucleome Informatics offers end-to-end mRNA Sequencing (mRNA-Seq) and data analysis services for a wide range of organisms including plants, animals and microbes.

mRNA-Seq enables comprehensive, quantitative profiling of the polyadenylated transcriptome and is a primary approach for studying gene expression, discovering differentially expressed genes and characterising transcript-level changes under experimental or environmental conditions.

Our analysis combines rigorous sequencing quality control, reference alignment or de novo transcriptome assembly, quantification, and differential expression analysis. Results are reported with functional annotations to support biological interpretation.

  • Sequencing data quality assessment and preprocessing
  • Reference genome alignment or de novo transcriptome assembly
  • Transcript and gene-level expression quantification
  • Differential expression (DE) analysis across experimental conditions
  • Functional annotation (GO, KEGG pathway enrichment)
  • Visualisation of DE results and expression patterns
  • Sample-wise and project-level QC and expression summaries
  • Secure delivery of raw, intermediate and final analysis files

Reference or De Novo

Reference-based analysis using HISAT2/STAR for organisms with a genome, or Trinity-based de novo assembly for non-model organisms lacking a reference.


Broad Organism Support

Analysis supported across kingdoms.

Human Animal Plant Microbial

Turn Around Time (TAT)

Varies by project

Turn around time varies based on the research objectives of the project, including sample number, sequencing platform and depth, and the scope of analysis.

Technology / Platforms

Short-read and long-read sequencing platforms, selected to match the study design.

Illumina NovaSeq 6000Short-read
PacBio RevioLong-read
PacBio Sequel IILong-read
Oxford Nanopore PromethION 2iLong-read
Learning Path

From first principles to expert-level analysis

New to RNA sequencing? Start with Foundations. Planning an experiment or reviewing results? Jump to Practitioner or Expert. Each level builds on the one before it, so you can read this page as a short course or dip in where you need to.

The big idea in plain language: genes are the instructions, and messenger RNA (mRNA) is the working copy a cell makes when it uses one. Counting mRNA molecules tells you which genes are switched on, and how strongly, in each sample.

LEVEL 1

Genes are copied into mRNA

Every cell carries the same DNA, but each cell type uses different genes. When a gene is used, its sequence is copied into mRNA, which is then read to make a protein.

LEVEL 1

mRNA-Seq counts the copies

RNA is converted to DNA, sequenced, and each read is matched back to its gene. A gene with more reads was producing more mRNA in that sample.

LEVEL 1

The poly(A) tail is a handle

Mature mRNA ends in a string of A bases. Capturing that tail pulls out mRNA and leaves behind ribosomal RNA, which makes up most of the RNA in a cell.

LEVEL 1

Reads become a count table

Reads are aligned to a genome (or assembled if there is none), then counted per gene. The result is a table of genes by samples: the starting point for every comparison.

LEVEL 1

Differential expression

Statistics compare groups, such as treated versus control, and list genes that are reliably higher or lower, with the size of the change (fold change) and a confidence value.

LEVEL 1

Pathways give the story

Hundreds of changed genes are hard to read. Enrichment analysis asks which biological processes, such as stress response or photosynthesis, are over-represented in the list.

For scientists planning a project or reading a report: how design and quality control decide whether a gene list can be trusted.

LEVEL 2

Replicates beat depth

Biological replicates measure natural variation, which statistics need. A large yeast study recommended at least six replicates per condition, and more when small changes matter (Schurch et al., 2016).

LEVEL 2

Choose the library to fit the sample

Poly(A) selection suits good-quality eukaryotic RNA. Ribosomal RNA depletion keeps non-polyadenylated RNA and tolerates some degradation. Strand-specific libraries keep track of which DNA strand was transcribed.

LEVEL 2

RNA quality shows in the reads

Degraded RNA gives reads piled at the 3′ end of genes. The RNA integrity number (RIN) is a pre-sequencing check; coverage plots are the post-sequencing check.

LEVEL 2

Raw counts are not comparable

A sample sequenced twice as deeply has about twice the counts. Normalization such as DESeq2's median-of-ratios or edgeR's TMM puts samples on one scale before testing.

LEVEL 2

Look at the samples first

PCA and a sample-correlation map reveal outliers, mislabelled samples and batch effects. These matter more than any single p-value.

LEVEL 2

Correct for thousands of tests

With about 20,000 genes tested, some look significant by chance. The false discovery rate (FDR) controls the expected share of false hits in the list.

For method developers and reviewers: the statistical and biological caveats that decide whether an apparent result holds up.

LEVEL 3

Counts follow a negative binomial

Real replicate counts vary more than a Poisson allows. DESeq2 and edgeR model this extra spread (dispersion) and share information across genes so that small studies remain usable (Love et al., 2014).

LEVEL 3

Model batch, don't just delete it

If samples were prepared in batches, include batch as a term in the design so the comparison is made within batches. This only works when batches are balanced across groups.

LEVEL 3

Fold changes of low-count genes are noisy

A gene with a handful of reads can show a huge log fold change by chance. Shrinkage estimators such as apeglm pull these toward zero so that rankings are more reliable (Zhu et al., 2019).

LEVEL 3

Transcript-level estimates help gene-level tests

Quasi-mapping tools such as Salmon estimate transcript abundance while correcting for biases, and tximport summarises them to genes with length offsets (Patro et al., 2017; Soneson et al., 2015).

LEVEL 3

Enrichment has built-in biases

Long genes have more reads and so more power to be called significant, which can inflate certain GO terms. The background should be the genes that were expressed, not the whole genome (Young et al., 2010).

LEVEL 3

De novo assemblies need their own QC

Without a genome, quality is judged by gene-space completeness (BUSCO), read back-mapping, contig statistics and redundancy, because many assembled transcripts are isoforms or fragments of the same gene.

Glossary of key terms
Read
One sequenced fragment of cDNA, stored in a FASTQ file with base-quality scores.
Poly(A) tail
A run of adenines added to the 3′ end of mature mRNA; the basis of mRNA capture.
rRNA
Ribosomal RNA: the most abundant RNA in cells, usually removed or avoided.
RIN
RNA integrity number, from 1 (degraded) to 10 (intact), measured before sequencing.
Alignment
Placing each read at its position in the genome, allowing for splicing across introns.
Count
The number of reads assigned to a gene or transcript in a sample.
TPM
Transcripts per million: a within-sample abundance unit corrected for gene length.
Size factor
A per-sample scaling value that corrects for sequencing depth and composition.
Dispersion
How much replicate counts vary beyond Poisson noise, for a given gene.
log2FC
Fold change on a log scale: +1 means doubled, −1 means halved.
FDR
False discovery rate: the expected share of reported hits that are false.
Isoform
One of several transcript variants produced from a single gene by splicing.
The Science

How a cell's RNA becomes a table of counts

These schematics show what mRNA is, how it is isolated and read, and how reads turn into numbers. Plots marked simulated are generated for this page, not client data.

Figures
Figure 1 From gene to captured mRNA schematic
1 · GENE IN DNA exon 1intronexon 2intronexon 3intronexon 4 transcription, then splicing joins the exons 2 · MATURE mRNA AAAAAAAAAA 5′ cap5′ UTRcoding sequence (makes the protein)3′ UTRpoly(A) tail dT oligo(dT) bead captures the tail Not captured: ribosomal RNA, transfer RNA The majority of cellular RNA lacks a poly(A) tail and is washed away.
Figure 1 | Why poly(A) capture enriches mRNA. Introns are removed by splicing, a cap and a poly(A) tail are added, and the tail lets magnetic oligo(dT) beads pull mature mRNA out of total RNA. Bacteria do not polyadenylate most mRNA, so microbial projects usually use ribosomal RNA depletion instead.
Figure 2 From RNA sample to sequencing reads infographic
STEP 01

RNA extraction

Total RNA is purified from tissue, cells or microbial culture, then checked for quantity, purity and integrity.

total RNA
RIN · A260/280
STEP 02

Select the RNA of interest

Poly(A) capture keeps mRNA; alternatively ribosomal RNA is depleted. Pre-enriched RNA can be supplied directly.

poly(A) capturerRNA depletion
enriched RNA
rRNA carry-over
STEP 03

Build the library

RNA is fragmented, copied into cDNA and given adapters and sample indexes so many samples can run together.

indexed cDNA library
size · yield
STEP 04

Sequence

Millions of fragments are read in parallel, from one end or both ends, producing FASTQ files.

FASTQ
Q30 · cluster density
Figure 2 | Typical wet-lab path. Kits, read length and depth are agreed for each project. Analysis-only projects supply FASTQ files and enter the workflow at the data QC step.
Figure 3 How reads become counts schematic
REFERENCE GENOME exon 1exon 2exon 3intronintron ALIGNED READS orange = read inside one exon · teal = spliced read crossing an intron Count table GeneSample 1Sample 2 Gene A1341 Gene B220198 Gene C02 Gene D1,5021,377 numbers are illustrative
Figure 3 | Aligning and counting. A splice-aware aligner (HISAT2 or STAR) places reads on the genome, including reads that jump across introns. A counter such as featureCounts then tallies reads per gene. Quasi-mapping tools (Salmon, Kallisto) skip the genome alignment and estimate transcript abundance directly.
Figure 4 Read coverage along a transcript simulated
Figure 4 | Degradation leaves a 3′ bias. Normalized coverage from the 5′ to the 3′ end of a gene. Intact RNA gives broadly even coverage; degraded poly(A)-selected RNA loses the 5′ end, which distorts expression estimates for long genes (Schroeder et al., 2006 for RIN).
Table 1 Poly(A) selection or rRNA depletion?
OptionStrengthsLimits
Poly(A) captureEfficient mRNA enrichment; fewer wasted reads; standard for plants, animals, humanNeeds fairly intact RNA; misses non-polyadenylated RNA; unsuitable for most bacteria
rRNA depletionKeeps non-coding and non-polyadenylated RNA; works with partly degraded RNA; used for bacteriaMore residual rRNA and intronic reads; species-specific probes
Tell us your sample and question. Library type is agreed per project. For analysis-only projects, check which method generated your data, because it changes what the counts mean.
UnitUse it for
Raw countsInput to DESeq2 and edgeR (they normalize internally)
Normalized countsComparing one gene across samples
TPMComparing genes within one sample
FPKM / RPKMLegacy; avoid for cross-sample comparison
Analysis Workflow

From raw reads to functional insight

Each sample passes through a quality-controlled analysis workflow designed to move from raw sequencing data to differential expression results and functional annotation.

Figure 5 Quality-gated RNA-Seq workflow infographic
STAGE 01

Data QC & pre-processing

Per-base quality, adapter content and read counts are reviewed; adapters are trimmed and low-quality bases filtered.

STAGE 02

Alignment or assembly

Splice-aware alignment to a reference, or Trinity assembly when no genome exists.

STAGE 03

Quantification

Reads are counted per gene or per transcript to build expression matrices.

STAGE 04

Sample-level QC

PCA and sample similarity reveal outliers and batch effects before any testing.

STAGE 05

Differential expression

DESeq2 and/or edgeR test each comparison, in pairwise, multi-group or multi-factor designs.

STAGE 06

Functional annotation

DE genes are annotated and tested for GO term and KEGG pathway enrichment.

STAGE 07

Visualisation

PCA plots, heatmaps, volcano plots and expression barplots of key genes.

STAGE 08

Reporting & delivery

QC, expression statistics, DE results and annotated outputs delivered through secure storage.

Figure 5 | Workflow. Metrics are reviewed before the next stage. Samples with unusual quality, alignment or replicate agreement are flagged to you rather than silently carried forward.
Design, QC & Normalization

Good statistics start before the sequencer

How many replicates, how deep to sequence, and whether the data are clean decide what can be concluded. All plots below use simulated data to show what each check looks like.

Figures
Figure 6 Replicates and statistical power interactive · model
Figure 6 | More replicates, more power. Approximate chance of detecting a gene at a strict significance level, for 1.5-, 2- and 3-fold changes, from a simple negative-binomial model. Variability and low counts both demand more replicates. A large empirical study recommended at least six per condition (Schurch et al., 2016).
Figure 7 Sequencing depth and genes detected simulated
Figure 7 | Diminishing returns. Genes with at least 10 reads as sequencing depth grows. Curves flatten once most expressed genes are covered; rare transcripts and isoforms keep needing depth. Broader expression ranges need more reads to saturate.
Figure 8 Base quality along the reads simulated
Figure 8 | Read 2 degrades first. Mean Phred quality per position for paired 150 bp reads. Shaded bands show spread. Low-quality tails and adapter sequence are trimmed before alignment; Q30 means about 1 error in 1,000 bases.
Figure 9 Where reads end up illustrative
Raw read pairsafter demultiplexing
100%
After trimmingadapters, low quality
98.5%
Alignedto the reference genome
96%
Uniquely alignedone best location
90%
Assigned to genescounted in the matrix
78%
Figure 9 | Read tracking. Illustrative retention for one good-quality sample. Low alignment suggests contamination, the wrong reference or poor RNA; a low assigned fraction suggests lots of intronic, intergenic or rRNA reads.
Figure 10 Sample-to-sample similarity simulated
Figure 10 | Do replicates agree? Pearson correlation of log-scaled expression between all pairs of 12 samples (three groups of four). Replicates should be more similar to each other than to other groups. A sample that does not fit is an outlier or may be mislabelled.
Figure 11 PCA and batch effects interactive · simulated
Figure 11 | Group or batch? Colour is the condition and shape is the preparation batch. Before adjustment, samples can separate by batch. Fixing this properly means putting batch in the model for testing; the adjusted view is for visualisation only (Leek et al., 2010).
Figure 12 Normalizing for library size simulated
Figure 12 | Putting samples on one scale. Boxplots of log-scaled counts for 12 samples. Left: raw counts differ with sequencing depth. Right: after dividing by size factors (the median-of-ratios method of DESeq2; TMM in edgeR does the equivalent) the medians line up and genes can be compared across samples (Love et al., 2014; Robinson and Oshlack, 2010).
Differential Expression

Finding genes that really change

Tests ask whether a gene's counts differ between groups by more than natural variation allows. Diagnostics show whether the statistical model fits, and effect size is read alongside confidence. Plots use simulated negative-binomial counts and a simplified Wald test, not the output of DESeq2 itself.

Figures
Figure 13 Mean-dispersion relationship simulated
Figure 13 | Borrowing strength across genes. Each grey dot is a gene's own noisy dispersion estimate from four replicates. The red curve is the trend across all genes and blue dots are final values pulled toward it. This sharing is why DESeq2 and edgeR work with few replicates (Love et al., 2014).
Figure 14 P-value histogram diagnostic simulated
Figure 14 | A healthy test. Genes with no change give a flat background; real effects add a spike near zero. The dashed line is the expected background level. A U-shape, a hill in the middle or a missing spike points to a model, batch or filtering problem.
Figure 15 MA plots with and without fold-change shrinkage simulated
Figure 15 | Taming noisy low-count genes. Each point is a gene; x is mean expression, y is log2 fold change, and coloured points are significant at FDR < 0.05. Left: raw estimates fan out at low counts. Right: shrinkage pulls uncertain estimates toward zero while keeping well-measured large changes (Zhu et al., 2019). Shrinkage is shown as context for ranking and plotting.
Figure 16 Volcano plot interactive · simulated
Figure 16 | Effect size versus confidence. Each point is a gene. Orange is higher and green lower in the treated group. The page's standard criterion is FDR < 0.05; adding a fold-change cut-off is optional and trades sensitivity for biological relevance.
Figure 17 Heatmap of top DE genes simulated
Figure 17 | Expression patterns. Rows are the 40 most significant genes and columns are samples grouped by condition. Colour is the z-score across samples (orange high, teal low). Consistent blocks across replicates show reproducible changes.
Table 2 Choosing a differential expression tool
ToolModel and normalizationGood for
DESeq2Negative-binomial GLM; median-of-ratios; shrunken dispersion and fold changeGeneral use, small to moderate replicate numbers, multi-factor designs
edgeRNegative-binomial GLM; TMM normalization; quasi-likelihood testsGeneral use; flexible contrasts; very similar results to DESeq2 for most data
limma-voomWeighted linear models on log-CPMLarge sample numbers and complex designs
Tools mostly agree. In a comparison of replicate-rich yeast data, edgeR and DESeq2 gave the best balance of true and false positives when fewer than 12 replicates were used (Schurch et al., 2016). Genes found by both tools are the most robust. This service reports DESeq2 and/or edgeR results.
Functional Enrichment

From gene lists to biological processes

Enrichment asks which functions occur more often in your gene list than chance would predict. It is a hypothesis generator, and its result depends heavily on the gene list and the background chosen.

Figures
Figure 18 GO and KEGG enrichment dot plot simulated
Figure 18 | Over-represented terms. Gene ratio is the share of the list belonging to a term; dot size is the gene count and colour the adjusted p-value. Term names are illustrative. Gene ontology and KEGG are the annotation sets used in this service.
Figure 19 Rank-based enrichment (GSEA-style) simulated · context
Figure 19 | No cut-off needed. All genes are ranked by change; the curve rises when members of a gene set cluster near the top (marks below the curve). It uses subtle coordinated shifts that a significance cut-off would miss (Subramanian et al., 2005). Shown for context; the standard deliverable is GO and KEGG enrichment of DE gene lists.
Table 3 Two ways to test a gene set
ApproachInputWatch out for
Over-representation (ORA)A DE gene list and a backgroundDepends on the cut-off; background must be expressed genes
Rank-based (GSEA)Every gene, ranked by a statisticNeeds a good ranking metric; permutation choices
Table 4 Pitfalls in enrichment results
PitfallSafeguard
Whole-genome backgroundUse genes detected in the experiment
Gene-length biasLength-aware methods such as goseq (Young et al., 2010)
Redundant GO termsSummarise related terms; focus on effect and gene counts
Incomplete annotationRemember non-model organisms have fewer annotated genes
De Novo Assembly

When there is no reference genome

For non-model organisms, transcripts are rebuilt from the reads themselves. The assembly then serves as the reference for quantification and differential expression.

Figures
Figure 20 How Trinity assembles transcripts schematic
Sequencing reads (no genome) 1 · Inchwormcuts reads into k-mers and greedily extends them into linear contigs 2 · Chrysalisgroups related contigs and builds a de Bruijn graph for each group 3 · Butterflyfollows read-supported paths through each graph to report isoforms
Figure 20 | Three modules. Reads are cut into overlapping words (k-mers), assembled into contigs, organised into graphs per gene, and resolved into full-length isoforms (Grabherr et al., 2011). Reads are later quasi-mapped back for quantification.
Figure 21 Contig length profile (Nx) simulated
Figure 21 | Assembly contiguity. Nx is the length such that contigs of at least that length hold x% of all assembled bases. Collapsing redundant isoforms (right-shifted curve) raises N50. Contiguity alone does not prove correctness, so completeness and read support are also checked.
Table 5 Judging a de novo transcriptome
CheckWhat it tells you
Read back-mapping rateShare of reads represented in the assembly; low values suggest missing or fragmented transcripts
BUSCO completenessPresence of near-universal single-copy genes for the lineage (Manni et al., 2021); high duplication hints at isoform redundancy
N50 and length distributionContiguity, best compared after redundancy reduction
Annotation rateShare of transcripts with hits in public databases; unannotated ones may be novel, noise or chimeric
Deliverables

Structured outputs for expression biology

At project completion, clients receive a structured set of sequencing, analysis and reporting outputs via secure shared storage.

Raw & Processed Data

FASTQ files — raw and trimmed sequence data.

QC Reports

Sample-level sequencing and preprocessing QC summaries.

Alignment Files

Sorted and indexed BAM files, where included in scope.

Quantification Matrices

Gene- or transcript-level count and expression matrices.

DE Results

Per-comparison tables of DE genes with statistics and fold changes.

Annotated Gene Lists

DE gene lists with functional annotation, GO and KEGG enrichment.

Summary Report

Sample-wise QC, alignment, quantification and DE analysis statistics.

Secure Delivery

Project files via secure shared storage or cloud access.

Sample Requirements

What we accept

Flexible input options for sequencing-plus-analysis or analysis-only projects.

Total RNA or Poly(A)-Enriched RNA

We accept total RNA or poly(A)-enriched RNA from biological samples including plant tissue, animal tissue, cells and other material, as well as pre-generated sequencing data where analysis-only services are required.

Project-specific requirements for RNA quantity, concentration, purity (A260/A280, A260/A230), integrity (RIN), sequencing depth, read length and replicate structure are finalised according to the study design and organism.

Contact us for a project-specific sample submission checklist and recommended sequencing depth.

Plant · Animal · Cells · Microbial
or pre-generated FASTQ

Aim for replicates firstUse biological, not technical, replicates. At least six per condition is a widely cited minimum for reliable gene lists (Schurch et al., 2016).
Check RNA integrityShare RIN and purity values (A260/A280, A260/A230) with the submission so library choice can be matched to the sample.
Randomise batchesSpread conditions across extraction and library batches. Record batch in the sample sheet so it can be modelled.
Keep handling identicalCollect, stabilise and store samples the same way in every group; handling differences can look like expression changes.
Case Studies & Performance

Proven across organisms

Published transcriptome studies on long-read (PacBio) and short-read (Illumina) platforms.

Indian white shrimp full-length transcriptome

Aquaculture species · PacBio Iso-Seq · Sequel II

PlatformPacBio Sequel II
Run8M SMRT cell, CCS
TissuesGills, hepatopancreas, muscle, post-larvae
Data generated238.98 Gb (120.7 M subreads)
Complete genes89.5% (vs 80.86% in tiger shrimp Iso-Seq)

The first comprehensive transcriptome for the species, providing a full-length reference for aquaculture breeding.

Katneni et al., Frontiers in Marine Science 2020. doi:10.3389/fmars.2020.605098
Nucleome Informatics is listed on the paper, with Dushyant Singh Baghel as co-author.

Onion bulb transcriptome for allergens and epitopes

Allium cepa · Illumina HiSeq 2000

PlatformIllumina HiSeq 2000
High-quality raw reads99,074,309 (~20 Gb)
Functional groups49
Use of the dataIgE epitope mapping, protein-structure prediction

The first transcriptome catalogue of onion bulb tissue, with genes grouped into 49 functional groups.

Rajkumar et al., PLOS ONE. doi:10.1371/journal.pone.0135387
The Frontier

What bulk mRNA-Seq tells you, and where transcriptomics is heading

Bulk mRNA-Seq remains the most cost-effective way to compare gene expression across conditions. Here is how it compares with newer approaches, and what recent research says about statistics and reproducibility.

Figure 22 Short reads versus full-length isoforms schematic
TWO ISOFORMS OF ONE GENE isoform B skips exon 2 Short reads (about 150 bp) most reads fit both isoforms; only junction reads discriminate Long read (full-length cDNA or RNA) a single read can span every exon (dark: isoform A, light: isoform B)
Figure 22 | Isoform ambiguity. Gene-level counts are robust, but telling isoforms apart from short reads relies on the few reads that cross a junction. Long-read platforms can sequence a whole transcript in one read, at higher cost per read (Amarasinghe et al., 2020).
Table 6 Choosing the right approach
ApproachStrengthsLimits
Bulk short-read mRNA-SeqCost-effective; mature tools; many samples and conditionsAverages over cell types; isoforms inferred
Long-read RNA-SeqFull-length isoforms, fusions, novel transcriptsHigher cost per read; fewer reads per run
Single-cell RNA-SeqCell-type resolution and rare populationsSparse; high cost; needs dissociation
Spatial transcriptomicsExpression in tissue contextLower depth; specialised platforms
Pick by question. To compare conditions across many samples, bulk mRNA-Seq is usually the efficient choice. If the answer depends on cell type or isoform, ask about extended designs.
BEST PRACTICE

A field-wide survey

A widely cited review covers each analysis step from experimental design to quantification, differential expression, splicing and functional analysis, and highlights the open challenges at each (Conesa et al., Genome Biology, 2016).

STATISTICS

Replicates drive reliability

Empirical benchmarking found that at least six biological replicates per condition are needed for reliable gene lists, and more when small fold changes matter (Schurch et al., RNA, 2016).

SINGLE CELL

Beware pseudoreplication

Single-cell differential expression methods that ignore variation between biological replicates can report hundreds of false discoveries. Treating samples, not cells, as the unit of replication is a safeguard (Squair et al., Nature Communications, 2021).

QUANTIFICATION

Bias-aware transcript estimates

Salmon models fragment-level biases when estimating transcript abundance, and tximport summarises these to gene level for DE tools (Patro et al., Nature Methods, 2017; Soneson et al., F1000Research, 2015).

REPRODUCIBILITY

Standardised pipelines

Community workflows such as nf-core/rnaseq package alignment, quantification and QC in versioned, containerised form so that analyses can be rerun identically (Ewels et al., Nature Biotechnology, 2020).

DECONVOLUTION

Estimating cell mixtures

Computational methods can estimate the proportions of cell types in a bulk sample from reference expression signatures, helping to interpret expression changes that reflect composition (Newman et al., Nature Methods, 2015).

Scope note. Nucleome's mRNA-Seq service covers the workflow and deliverables described on this page: data QC and preprocessing, reference alignment (HISAT2 or STAR) or Trinity de novo assembly, quantification (featureCounts, Salmon or Kallisto), differential expression with DESeq2 and/or edgeR for pairwise, multi-group and multi-factor designs, GO and KEGG enrichment, visualisation and secure delivery. Power estimates, fold-change shrinkage, rank-based enrichment, long-read and single-cell sequencing, deconvolution and splicing analysis are shown to place the service in its scientific context; contact our team to discuss extended or project-specific analyses.
Questions

Frequently asked, in plain language

From the basics to the finer points of design and statistics.

basicWhat is mRNA-Seq?
It sequences the messenger RNA in a sample and counts how many reads come from each gene. The counts are a measure of how active each gene was, and can be compared between conditions.
intermediateReference-based or de novo?
If your organism has a good genome, reference-based analysis with HISAT2 or STAR is preferred. For organisms with no genome, we assemble the transcriptome de novo with Trinity and use it as the reference.
intermediateHow many replicates do I need?
More is better, and replicates matter more than depth. A large benchmarking study recommended at least six per condition, and more when small changes matter. Our team agrees replicate numbers and sequencing depth with you for each design.
intermediateDoes RNA quality matter?
Yes. Degraded RNA biases coverage toward the 3′ end and can distort results, especially after poly(A) selection. Share RIN and purity values with your submission so we can match the library method to the sample.
intermediateWhy DESeq2 and edgeR?
Both model counts with a negative binomial distribution, share information across genes, and perform well with modest replicate numbers. Genes called by both are the most robust.
advancedHow is batch handled?
Record batch in the sample sheet. When batches are balanced across conditions, it can be included as a term in the multi-factor design so comparisons are made within batches. If batch and condition are fully confounded, no statistical method can separate them.
advancedWhy are some GO results misleading?
Enrichment depends on the gene list, the annotation and the background set, and can be biased by gene length. Treat enriched terms as hypotheses and prefer a background of genes that were expressed in the experiment.
advancedCan mRNA-Seq tell me about isoforms or cell types?
Only partly. Short-read bulk data give gene-level answers reliably; isoforms are harder to resolve, and cell-type information is averaged. Long-read, single-cell or deconvolution approaches address these questions and can be discussed at the design stage.
References

Key literature & resources

Selected sources behind the methods and reference values on this page. Figures labelled "simulated", "schematic" or "illustrative" are generated for this page, not client data.

Show all 30 references
  1. Conesa A, et al. A survey of best practices for RNA-seq data analysis. Genome Biology 2016;17:13. Record
  2. Schurch NJ, et al. How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use? RNA 2016;22:839–851. Abstract
  3. Stark R, Grzelak M, Hadfield J. RNA sequencing: the teenage years. Nature Reviews Genetics 2019;20:631–656.
  4. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology 2014;15:550. PDF
  5. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 2010;26:139–140.
  6. Robinson MD, Oshlack A. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biology 2010;11:R25.
  7. Law CW, et al. voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biology 2014;15:R29.
  8. Zhu A, Ibrahim JG, Love MI. Heavy-tailed prior distributions for sequence count data: removing the noise and preserving large differences. Bioinformatics 2019;35:2084–2092.
  9. Dobin A, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013;29:15–21.
  10. Kim D, Paggi JM, Park C, Bennett C, Salzberg SL. Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype. Nature Biotechnology 2019;37:907–915.
  11. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 2014;30:923–930.
  12. Patro R, et al. Salmon provides fast and bias-aware quantification of transcript expression. Nature Methods 2017;14:417–419. DOI
  13. Bray NL, et al. Near-optimal probabilistic RNA-seq quantification. Nature Biotechnology 2016;34:525–527.
  14. Soneson C, Love MI, Robinson MD. Differential analyses for RNA-seq: transcript-level estimates improve gene-level inferences. F1000Research 2015;4:1521.
  15. Grabherr MG, et al. Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nature Biotechnology 2011;29:644–652.
  16. Manni M, et al. BUSCO update: novel and streamlined workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic, and viral genomes. Molecular Biology and Evolution 2021;38:4647–4654.
  17. Schroeder A, et al. The RIN: an RNA integrity number for assigning integrity values to RNA measurements. BMC Molecular Biology 2006;7:3.
  18. Leek JT, et al. Tackling the widespread and critical impact of batch effects in high-throughput data. Nature Reviews Genetics 2010;11:733–739.
  19. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society B 1995;57:289–300.
  20. Subramanian A, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. PNAS 2005;102:15545–15550.
  21. Young MD, et al. Gene ontology analysis for RNA-seq: accounting for selection bias. Genome Biology 2010;11:R14.
  22. Wu T, et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. The Innovation 2021;2:100141.
  23. Ashburner M, et al. Gene Ontology: tool for the unification of biology. Nature Genetics 2000;25:25–29.
  24. Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Research 2000;28:27–30.
  25. Amarasinghe SL, et al. Opportunities and challenges in long-read sequencing data analysis. Genome Biology 2020;21:30.
  26. Squair JW, et al. Confronting false discoveries in single-cell differential expression. Nature Communications 2021;12:5692. Record
  27. Newman AM, et al. Robust enumeration of cell subsets from tissue expression profiles. Nature Methods 2015;12:453–457.
  28. Ewels PA, et al. The nf-core framework for community-curated bioinformatics pipelines. Nature Biotechnology 2020;38:276–278.
  29. Katneni et al. Indian white shrimp full-length transcriptome (PacBio Iso-Seq). Frontiers in Marine Science 2020. doi:10.3389/fmars.2020.605098
  30. Rajkumar et al. Onion bulb transcriptome for allergens and epitopes (Illumina). PLOS ONE. doi:10.1371/journal.pone.0135387