- Gene and transcript abundance measures are used in bioinformatics to estimate the expression levels of genes and their individual transcripts in a biological sample. They provide quantitative information about how actively genes are expressed and how much RNA is produced under particular biological conditions. These measurements are fundamental to transcriptomics and RNA sequencing (RNA-seq), enabling researchers to compare gene expression across tissues, developmental stages, environmental conditions, disease states, and experimental treatments. Gene abundance refers to the overall expression of a gene, whereas transcript abundance measures the expression of individual RNA transcripts or isoforms produced from that gene.
- A single gene can produce multiple transcript isoforms through alternative splicing, alternative transcription start sites, alternative polyadenylation, and other RNA-processing mechanisms. Consequently, gene-level abundance generally combines expression information from all transcripts associated with a gene, while transcript-level abundance attempts to quantify each isoform separately. This distinction is important because different transcript isoforms may encode proteins with different structures and functions, produce noncoding RNAs, or undergo different regulatory processes. Changes in transcript abundance can therefore occur even when the total abundance of the corresponding gene remains relatively unchanged.
- In RNA-seq experiments, abundance is estimated from sequencing reads or fragments that originate from RNA molecules. After RNA extraction, library preparation, sequencing, quality control, and alignment or transcript quantification, reads are assigned or probabilistically allocated to genes and transcripts. Gene-level counting tools such as featureCounts and HTSeq-count estimate the number of reads overlapping annotated genes. Transcript-level quantification tools such as Salmon, Kallisto, and RSEM estimate the abundance of individual transcripts, including cases where reads cannot be uniquely assigned to a particular isoform. These tools use reference genome or transcriptome annotations, including gene models, exon structures, transcript sequences, and genomic coordinates.
- The simplest expression measurement is the raw read count, which represents the number of sequencing reads or fragments assigned to a gene or transcript. Raw counts are widely used in differential gene expression analysis because they retain the count-based information required by statistical models. However, raw counts cannot be directly compared across genes or samples without considering differences in sequencing depth, transcript length, and other technical factors. A highly expressed long transcript may naturally receive more reads than a short transcript with the same number of RNA molecules. Similarly, a sample with greater sequencing depth may produce more reads for nearly all expressed genes.
- To account for transcript or gene length and sequencing depth, several normalized abundance measures have been developed. Reads per kilobase per million mapped reads (RPKM) adjusts read counts for gene or transcript length and the total number of mapped reads. It was initially developed for single-end RNA sequencing. Its formula is:
- RPKM = (C * 109) / (N * L)
- Where:
- C = number of reads assigned to a gene or transcript.
- N = total number of mapped reads.
- L = length of the gene or transcript in bases.
- Fragments per kilobase per million mapped fragments (FPKM) provides a similar normalization for paired-end sequencing. In paired-end sequencing, two reads are generated from opposite ends of the same RNA-derived fragment. FPKM counts fragments rather than treating the two reads as independent observations. Its formula is:
- FPKM = (C * 109) / (F * L)
- Where:
- C = number of fragments assigned to a gene or transcript.
- F = total number of mapped fragments.
- L = length of the gene or transcript in bases.
- Both RPKM and FPKM account for feature length and sequencing depth. However, they have limitations when comparing expression levels across samples because their values can be influenced by differences in transcriptome composition. For this reason, transcripts per million (TPM) is widely used for describing relative transcript abundance.
- Transcripts per million (TPM) first divides the read or fragment count for each transcript by its effective length. These length-normalized values are then divided by their sum across all quantified transcripts and multiplied by one million. The formula is:
- TPMi = (Ci / Li) / SUM(Cj / Lj) * 106
- Where:
- Ci = number of reads or fragments assigned to transcript i.
- Li = effective length of transcript i.
- j = index representing each transcript.
- n = total number of quantified transcripts.
- The denominator represents the sum of length-normalized counts across all quantified transcripts.
- The effective length accounts for the influence of transcript length on the probability of obtaining sequencing fragments from that transcript. TPM has an important mathematical property: the TPM values of all quantified transcripts in a sample sum to one million. This makes TPM useful for describing and comparing relative transcript abundance profiles. However, TPM values are relative measurements rather than absolute numbers of RNA molecules. They can also be affected by changes in the overall composition of the transcriptome.
- Counts per million (CPM) is another abundance measure that expresses the number of reads assigned to a gene or transcript per million mapped reads. It adjusts for differences in sequencing depth but does not correct for gene or transcript length. Its formula is:
- CPM = (C / N) * 106
- Where
- C represents the number of reads or fragments assigned to a feature and N represents the total number of mapped reads or fragments, according to the counting convention used.
- CPM is useful for filtering lowly expressed genes, examining expression patterns, and performing exploratory analyses. Log counts per million (log-CPM) are frequently used for visualization and statistical modelling because logarithmic transformation reduces the influence of extremely high expression values and makes expression distributions easier to examine. A commonly used form with a small prior count is: logCPM =
- LOG2(((C + a) / (N + 2*a)) * 106)
- Where
- a is a small prior count used to avoid taking the logarithm of zero. The exact calculation and adjustment can vary between software packages.
- Although RPKM, FPKM, TPM, and CPM are useful for describing expression abundance, they should not be treated as interchangeable measurements. RPKM and FPKM normalize for feature length and sequencing depth, TPM normalizes length-adjusted abundance so that the total within a sample equals one million, and CPM adjusts for sequencing depth without correcting for feature length. The choice of measure depends on the research objective and the type of comparison being performed.
- For differential gene expression analysis, raw counts are generally preferred over RPKM, FPKM, or TPM as input to statistical methods such as DESeq2, edgeR, and limma-voom. These methods apply their own normalization procedures to account for differences in library size and RNA composition. DESeq2 commonly uses median-of-ratios normalization, while edgeR uses the trimmed mean of M-values (TMM). Limma-voom transforms count data into log-counts per million and models the relationship between mean expression and observational variance. These approaches help produce statistically reliable comparisons between experimental groups.
- Gene and transcript abundance estimation is complicated by overlapping genes, shared exons, repetitive sequences, and transcript isoforms with similar or identical regions. Reads originating from these regions may map to multiple genomic locations or transcripts, making their assignment uncertain. Computational tools such as Salmon, Kallisto, and RSEM use probabilistic approaches to estimate transcript abundance from ambiguous reads. Transcript-level estimates can subsequently be summarized into gene-level measurements using tools such as tximport, which imports transcript abundance estimates into R-based differential expression workflows. The accuracy of these estimates depends on sequencing depth, read length, transcriptome annotation quality, and the complexity of transcript structures.
- Long-read sequencing technologies, including Oxford Nanopore and PacBio sequencing, provide additional opportunities for transcript abundance estimation. Their longer reads can span multiple exons and identify complete transcript structures, helping distinguish isoforms that are difficult to resolve using short reads. Tools such as FLAIR, TALON, StringTie2, and bambu support transcript identification and quantification using long-read data. Combining short-read and long-read sequencing can improve both abundance estimation and transcript structure characterization.
- Gene and transcript abundance can also be measured using experimental techniques such as quantitative PCR (qPCR), reverse-transcription quantitative PCR (RT-qPCR), and digital PCR. RT-qPCR measures the abundance of selected RNA targets through reverse transcription and amplification with fluorescence detection. Digital PCR partitions a sample into many individual reactions and estimates target concentration from the proportion of positive and negative partitions. External RNA spike-ins, synthetic RNA standards, and calibrated reference materials can help estimate absolute transcript abundance or assess technical variation. However, absolute quantification requires careful calibration and consideration of RNA extraction efficiency, sample input, and molecular losses.
- Gene and transcript abundance measurements are essential for identifying differentially expressed genes, studying tissue-specific expression, detecting alternative splicing, examining developmental regulation, and investigating molecular responses to environmental conditions. In animal breeding and livestock genomics, these measurements help identify genes and transcript isoforms associated with growth, feed efficiency, milk production, meat quality, fertility, disease resistance, immune responses, and adaptation to environmental stress. Expression profiling can also support the identification of candidate genes and biological pathways involved in economically important traits.
- The interpretation of abundance measures requires attention to biological and technical variation. RNA integrity, sample composition, sequencing depth, library preparation, batch effects, reference genome quality, transcript annotation, and normalization strategy can all influence estimated abundance. Differences in tissue cellular composition may also produce apparent expression changes even when expression within individual cell types remains unchanged. For this reason, gene and transcript abundance should be interpreted alongside experimental design, biological replication, quality-control results, and appropriate statistical analysis. No single abundance measure is suitable for every purpose: raw counts are generally appropriate for count-based differential expression analysis, TPM is useful for relative transcript abundance and expression profiling, and specialized calibrated methods are needed when absolute RNA quantities are required.