Adjusted p-value (padj)
A p-value corrected for the fact that thousands of genes are tested at once. It is the number you trust when deciding whether a gene's change is real. A common cutoff is below 0.05.
Alignment
Working out where each short sequencing read came from in a reference genome, like finding where a puzzle piece fits.
ANOVA
Analysis of variance: a test for whether the averages of three or more groups differ. The rank-based version for non-normal data is the Kruskal-Wallis test.
BAM
The compressed, computer-friendly version of a SAM file (aligned reads). Same information, much smaller.
Batch effect
A systematic technical difference between groups of samples that were processed at different times, by different people, or on different machines, with nothing to do with the biology. If ignored, it can look like a real result.
BED
A simple text format that lists regions of a genome (chromosome, start, end). Often used to mark features of interest.
bedtools
A fast command-line toolkit for genome arithmetic: finding overlaps between sets of regions, merging them, and similar interval operations on BED files.
Bioconductor
A large collection of R packages built specifically for biological data analysis. DESeq2 lives here.
Bioinformatics
Using computers and code to make sense of biological data such as DNA, RNA, and proteins.
BioMart
A query tool (part of Ensembl) for pulling and mapping gene, transcript, and protein IDs in bulk.
Biopython
A free Python toolkit for working with biological sequences and data.
BLAST
Basic Local Alignment Search Tool: the standard, fast way to search a huge database for sequences similar to your query. You judge a hit by its E-value, not just percent identity.
chi-square test
A test for whether counts across categories differ from what you would expect by chance. For small counts, use Fisher's exact test instead.
CIGAR
A short code in a SAM/BAM record describing how a read lines up to the reference: matches (M), insertions (I), deletions (D), soft-clips (S), and skipped regions (N).
Clustering
Grouping samples or genes by how similar they are, so patterns and relationships become visible.
conda
A tool that installs software and manages separate "environments". Popular in bioinformatics because it can install non-Python command-line tools too.
Confounding
When two effects are tangled together so you cannot tell them apart, for example if every treated sample was processed on a different day than the controls.
Count matrix
A table with one row per gene and one column per sample, where each cell is how many sequencing reads were counted for that gene. The starting point of most RNA-seq analyses.
CPM (counts per million)
Read counts scaled by library size so samples sequenced to different depths can be compared. It does not correct for gene length.
DESeq2
The standard R/Bioconductor package for finding genes that change between conditions (differential expression).
Design formula
A short recipe (like ~ condition) that tells DESeq2 which sample features to model and what comparison you care about.
Differential expression
Finding which genes are more or less active between two conditions, for example treated versus untreated cells.
Dispersion
A measure of how much a gene's counts bounce around between replicates of the same group. Noisy genes need a bigger change before they are called significant.
DNA
The molecule that stores the genetic instructions of a cell, written in four letters: A, T, G, C.
Effect size
How big a difference is, not just whether it is statistically significant. A tiny effect can be significant in a large study yet not matter in practice.
Ensembl
A major public database of genomes and gene annotations.
Entrez
NCBI's search and retrieval system (and its E-utilities API) for fetching sequences and records programmatically.
Environment (virtual environment)
A self-contained box holding a specific set of software packages, so one project cannot break another.
E-value (Expect value)
In a BLAST search, how many hits this good you would expect by chance in a database this size. Smaller is more trustworthy: 1e-50 is strong, 2 is meaningless.
FASTA
A simple text format for storing sequences: a header line starting with ">" followed by the sequence (DNA, RNA, or protein).
FASTQ
Like FASTA, but each base also carries a quality score showing how confident the sequencer was. The raw output of sequencing.
Fisher's exact test
A test for a small table of category counts (often 2x2), used when counts are too small for a chi-square test.
FLAG
A number in a SAM/BAM record that packs several yes/no facts about a read (paired, mapped, reverse strand, duplicate) into one value.
FPKM
Fragments per kilobase per million: an expression unit correcting for sequencing depth and gene length. TPM is usually preferred for comparing across samples.
Functional enrichment
Asking whether your list of changed genes is unusually full of genes with a shared role (a pathway or function), to learn what biology is changing.
Gene
A stretch of DNA that carries the instructions to make a product, usually a protein.
GenBank
NCBI's richly annotated sequence format and database.
Genome
The complete set of DNA in an organism.
Genome build
A specific version of a reference genome, such as GRCh37/hg19 or GRCh38/hg38. The same position has different coordinates in different builds, so always record which one you used.
GEO (Gene Expression Omnibus)
A public NCBI repository of gene-expression datasets you can download and reanalyze.
GFF / GTF
Text formats that describe where genes, exons, and other features sit on a genome (annotation).
GO (Gene Ontology)
A standardized vocabulary describing what genes do: their biological processes, molecular functions, and locations. Used in enrichment analysis.
GSEA (Gene Set Enrichment Analysis)
A method that scans your whole ranked gene list to see if a group of related genes trends up or down together.
Heatmap
A grid of coloured cells that shows patterns in data at a glance, for example which genes are high or low across samples.
Homology
Shared ancestry between sequences. It is an inference (yes or no), not a percentage; high sequence similarity is the evidence for it. "60% homologous" is a misnomer for "60% identical."
Index (genome index)
A precomputed lookup structure that lets an aligner find where a read belongs without scanning the whole genome, like the index at the back of a book.
KEGG
A database of biological pathways (chains of interacting genes and molecules). Used in enrichment analysis.
Kruskal-Wallis test
The rank-based alternative to ANOVA, for comparing three or more groups when the data is not roughly normal.
LFC shrinkage
Pulling unreliable fold changes (often from low-count genes) toward zero, so rankings and plots reflect well-measured effects.
liftOver
A tool that converts genomic coordinates from one genome build to another using a chain file.
Linear model (regression)
A model that describes an outcome as a straight-line function of one or more predictors. Many familiar tests, including the t-test and ANOVA, are special cases of it.
log2 fold change
How much a gene changed, on a log2 scale. +1 means it doubled, -1 means it halved. The sign is the direction, the size is the magnitude.
MA plot
A plot of fold change against average expression level. A health check that shows whether changes are spread fairly across genes.
MAPQ
Mapping quality: how confident the aligner is that a read is in the right place. A low value (near 0) often means the read maps equally well to several spots.
Median-of-ratios
The normalization DESeq2 uses on raw counts to correct for differences in sequencing depth between samples.
Multiple sequence alignment (MSA)
Lining up three or more sequences at once to reveal columns conserved across a whole family; conserved columns often mark functionally important residues. Tools: Clustal Omega, MUSCLE.
Multiple testing
The problem that testing thousands of genes at once produces many false positives by chance, which is why we adjust p-values.
NCBI
The US National Center for Biotechnology Information, home to many core biological databases.
NGS (Next-Generation Sequencing)
Modern technology that reads DNA or RNA in millions of short fragments at once.
Normalization
Adjusting counts so samples sequenced at different depths can be compared fairly, so differences reflect biology and not data volume.
p-value
The probability of seeing a difference this big by chance alone. Small means surprising. For many-gene tests, use the adjusted p-value instead.
pandas
A Python library for working with tables of data, like a programmable spreadsheet.
PCA (Principal Component Analysis)
A way to squeeze many measurements into two axes so you can see, at a glance, which samples group together.
Pearson correlation
A measure of how strongly two numeric variables move together in a straight line. For non-linear or ranked relationships, use Spearman.
pip
Python's package installer; it fetches and installs Python libraries.
Power (statistical power)
The chance a study detects a real effect if one exists. It rises with bigger effects, more samples, and less noise. Underpowered studies miss real effects and exaggerate the ones they report.
Reads
The short sequences a sequencing machine produces, each one a fragment of the original DNA or RNA.
Reference genome
A standard, agreed-upon genome sequence that new reads are compared against.
Replicate (biological vs technical)
A biological replicate is a separate sample (a different animal or culture); a technical replicate is the same sample measured again. Differential expression needs biological replicates.
RNA
A molecule transcribed from DNA. Measuring RNA tells you which genes a cell is actively using.
RNA-seq
Sequencing the RNA in a sample to measure how active each gene is.
SAM
A text format that stores aligned reads (where each read mapped on the genome). BAM is its compressed form.
Sequence
The order of letters (bases) in DNA or RNA, or of amino acids in a protein.
Single-cell RNA-seq
RNA-seq that measures gene activity in individual cells rather than averaged over millions.
Spearman correlation
A rank-based measure of whether two variables rise and fall together, robust to outliers and to non-linear but consistent relationships.
SRA (Sequence Read Archive)
A public NCBI repository of raw sequencing reads.
Substitution matrix
A scoring table (such as BLOSUM62 or PAM) for how favorable each amino-acid swap is, based on real related proteins. It makes protein alignment more sensitive than raw DNA matching.
t-test
A test for whether the averages of two groups differ. The rank-based version for small or skewed data is the Wilcoxon test.
TPM (transcripts per million)
An expression unit correcting for both sequencing depth and gene length that sums to the same total in every sample, which makes it good for comparing across samples. Do not feed TPM to DESeq2; it needs raw counts.
Transcription
The cell copying a gene's DNA into RNA, the first step in using a gene.
Transcriptomics
The study of all the RNA in a sample: which genes are on, off, up, or down.
UMI (unique molecular identifier)
A short random tag added to each molecule before PCR (common in single-cell RNA-seq) so duplicate copies can be collapsed and counted once.
UniProt
A major public database of protein sequences and information.
Variance-stabilizing transformation (vst)
A transform that puts genes on a comparable scale before plotting or clustering, so high-count genes do not dominate.
VCF (Variant Call Format)
A text format that lists differences (variants) between a sample's DNA and the reference genome.
venv
Python's built-in tool for creating virtual environments.
Volcano plot
A scatter plot of significance versus fold change. The genes that matter fly up into the top corners.
Wilcoxon test
The rank-based alternative to the t-test, for comparing two groups when the data is small or not normally distributed. Also called Mann-Whitney.
z-score
Rescaling each gene so a heatmap shows its relative change (up or down versus its own average) rather than its absolute level.
No term matches that search. Try another word, or send us a suggestion from the feedback button.