Bioinformatics
reference

A

Alignment

Matching reads to a reference sequence. BWA and Bowtie2 align short DNA reads from FASTQ. The result contains read coordinates and confidence scores. Repeats create ambiguity. Alignment alone does not call variants or quantify expression.

B

Batch effect

Systematic measurement differences between technical batches. A comparison model can account for them. With complete confounding, batch and condition effects cannot be separated. Corrections used for clustering are not always suitable for statistical tests.

Bulk transcriptomics

A measurement of the combined RNA profile of a sample. Differences reflect gene expression within cells and cell composition. Analysis requires counts, groups, donors, and technical batches. Library type determines which RNAs are measured.

C

Cell-free DNA

DNA outside cells. Plasma contains DNA fragments from different tissues. Tumor DNA is only a possible component. Variants can originate from blood cells. Sensitivity depends on the number of independent molecules and measurement errors. An absent signal does not prove that a tumor is absent.

Cell-free RNA

RNA outside cells. Its profile depends on source tissues, RNA degradation, and the extraction method. It does not directly reproduce a tissue transcriptome. Cell contamination, platelets, hemolysis, and processing delays can change the result.

Confounding

An association between the condition of interest and another factor that could explain the result. If all controls occupy one batch and all treated samples another, their effects cannot be separated. Correction cannot recover the missing comparison. Distributing conditions across batches reduces this risk.

Copy number

The number of copies of a DNA region in a cell. Copy-number changes are inferred from coverage depth and allele fractions. In tumors, estimates depend on tumor purity and ploidy. The output contains genomic segments with estimated copy numbers.

D

Deconvolution

Estimating cell-type contributions to a mixed RNA profile. Results depend on reference completeness and its match to the tissue. RNA fractions are not cell fractions. Estimating a spot's composition does not recover individual cell coordinates.

Differential expression

Estimating expression differences between conditions while accounting for variation between replicates. Results include effect size, uncertainty, and adjustment for multiple testing. Comparing donor groups requires accounting for variation between donors.

E

Effect size

The magnitude of a difference on a chosen scale. A log2 expression ratio of +1 means a twofold increase; -1 means a twofold decrease. Direction depends on the reference group. A large effect can have high uncertainty. Statistical significance does not guarantee practical value.

Epigenetics

Studying gene regulation without changes to the DNA sequence. DNA methylation, chromatin accessibility, and histone modifications require different measurements. An accessible chromatin region does not prove activity of the nearest gene.

F

FASTQ

A standard text format for sequencing reads. A read usually occupies four lines: identifier, sequence, separator, and base quality scores. FASTQ has no genomic coordinates; alignment determines them. A sample can span multiple files.

G

Gene set analysis

Testing for a shared signal among genes with a common function. Results depend on the gene set and the genes tested. Overlapping sets can reflect one signal. RNA enrichment alone does not prove pathway activation.

GeoMx (RNA)

A spatial transcriptomics method for selected tissue regions. DNA barcodes attached to RNA probes are collected from a region and sequenced. Outputs include counts by region and metadata. A profile usually mixes several cells. The probe set limits the genes measured. Region selection determines the comparison.

I

Illumina sequencing

The most widely used short-read NGS technology. The instrument determines sequences from signals produced during DNA synthesis. Signal processing produces FASTQ. Analysis depends on read length, paired-end status, and the indexing scheme.

P

Pseudobulk

Summing raw counts for cells of one type within a sample. Profiles are formed separately for each donor and condition. This preserves replicates for group comparisons. Pooling donors loses this information. An absent cell type does not mean zero expression of all its genes.

S

Single-cell transcriptomics

Measuring RNA profiles of individual cells or nuclei. It distinguishes cell composition and cell states. Cell and nuclear profiles differ. A zero count does not prove absent expression. Cells from one donor do not replace replicates. Distance in a two-dimensional projection is not a measure of biological difference.

Spatial transcriptomics

Measuring RNA while preserving its position in tissue. Analysis requires expression, coordinates, scale, and image alignment. A spot and a cell are different units. Cell profiles depend on segmentation. Proximity does not prove interaction. Spots from one section do not replace independent samples.

T

Transcriptomics

Measuring RNA composition and abundance. Bulk analysis describes cell mixtures; single-cell analysis describes individual cells. Spatial analysis preserves tissue position; cell-free RNA analysis describes extracellular mixtures. The method determines which conclusions are supported. RNA abundance is not protein abundance or process activity.

V

Visium

A spatial transcriptomics method that uses sequencing and positional barcodes. Outputs include counts by spot or bin, coordinates, and a tissue image. A spot can contain several cells. In Visium HD, bin size affects signal and resolution.

W

Whole-exome sequencing

Sequencing mainly the protein-coding regions of the genome. It supports variant detection in exons. Analysis requires capture intervals and coverage assessment for each region. An absent variant in a poorly covered region is uninformative. Changes outside exons require other approaches, such as whole-genome sequencing.

Whole-genome sequencing

Sequencing the genome without selecting only coding regions. It provides data for detecting point variants, rearrangements, and copy-number changes. Coverage remains uneven. Repeats and complex regions limit variant detection. More data does not guarantee complete genome reconstruction.

X

Xenium

A spatial transcriptomics method that detects selected RNAs directly in tissue through imaging. Outputs include molecule coordinates, images, and cell counts. Panel content limits the available genes. Cell boundaries affect molecule assignment and expression estimates.