Bioinformatics Glossary

原文：http://homepages.ulb.ac.be/~dgonze/TEACHING/bioinfo_glossary.html

Affine gap costs: A scoring system for gaps within alignments that charges a penalty for the existence of a gap and an additional per-residue penalty proportional to the gaps length. 

Algorithm: A fixed procedure, implemented in a computer program. 

Alignment score: A numerical value that describes the overall quality of an alignment. Higher numbers correspond to higher similarity, which is unlikely to have been obtained by chance. 

Bit score: A log-scaled version of a score. 

BLAST: (Basic Local Alignment Search Tool). A heuristic sequence comparison algorithm, developed at the National Center for Biotechnology Information (NCBI), that is used to search sequence databases for optimal local alignments to a query sequence. 

BLOSUM (BLOck SUbstitution Matrix): A collection of substitution matrices based on conserved protein domains. 

Bootstrapping: A statistical method often used to estimate the reproducibility of specific features of clustering or phylogenetic trees. 

Cluster analysis: (or clusering): A process of assigning data points (sequences) into groups (clusters). 

Command line: Interacting with software by typing specific commands. Generally considered less user friendly than a graphical user interface. 

Comparative genomics: The study of comparing complete genome sequences, often by computational methods, to understand general principles of genome structure and function. 

Controlled vocabulary: A vocabulary that contains specific words that are consistently applied to all entries in a database. Gene ontology (GO) or MeSH system are examples of controlled vocabulary. 

DNA chip technology: New technology for parallel processing thousands of DNA segments, such as for detecting mutation patterns in genomic DNAs or expression patterns of mRNAs. 

Dynamic programing: A type of algorithm widely used for constructing sequence alignments, which guaranty to return the optimal solution. 

E-value (Expectation value): correction of the p-value for multiple testing. In the context of database searches, the e-value is the number of distinct alignments, with score equivalent to or better than the one of interest, that are expected to occur in a database search purely by chance. The lower the E value, the more significant the score is. 

EST (Expressed Sequence Tag): A short cDNA (complementary DNA) sequence from an expressed gene and which is assumed to long enough to be specific to a given gene. Often used to confirm gene prediction. 

Extreme value distribution (EVD): The probability distribution applicable to the scores of optimal local alignments. EVD is used to compute the p-value and the e-value when a database search is performed. 

False negative (often denoted FN): something predicted as negative, but which is actually positive. Ex: a gene which is not predicted to be regulated by a transcription factor X, although in reality it is regulated by X. See also "true positive" and "false positive". 

False positive (often denoted FP): something predicted as positive, but which is actually not. Ex: a gene which is predicted to be regulated by a transcription factor X, but which is in reality not regulated by X. See also "true positive" and "false negative". 

FASTA: A popular heuristic sequence comparison algorithm (Pearson & Lipman), that is used to search sequence databases for optimal global alignments to a query. 

Filtering (in BLAST): See masking. 

Free-form text The opposite of a "controlled vocabulary". Free text has no structured set of words, such that two related entries might not be identified in a search because different words are used to describe each entry. 

ftp (File Transfer Protocol): A method for transferring files across a network. 

Functional genomics The study of predicting gene function using genomic information, and, in a broader sense, obtaining an overall picture of genome functions, including the expression profiles at the mRNA level (transcriptome) and the protein level (proteome). 

Gap: Within an alignment of two sequences, several adjacent null characters in one sequence aligned with adjacent letters in the other. 

Gap score (or gap cost): The score (or cost) assigned to a gap in an alignment (linear or affine). 

Gapped alignment: An alignment in which gaps are permitted. 

GenBank: A data bank of genetic sequences developped at National Institutes of Health (NIH). 

Gene family: Two or more genes that are related by divergent evolution from a common ancestor, either by speciation or gene duplication. 

Gene fusion: A fusion gene is a hybrid gene formed from two or several previously separate genes. This can lead for example to a situation where one gene in an organism (i.e. yeast) corresponds to two or several genes in a other species (e.g. e. coli). 

Global alignment: The alignment of two (or more) complete nucleic acid or protein sequences. 

Graphical user interface Software that allows a user to interact via user-friendly menu and mouse-driven commands, as is typical of Macintosh and Windows applications and less common for UNIX applications; as opposed to a "command line" interface of typed or scripted commands. 

Heuristics A term in computer science that refers to "guesses" made by a program to obtain approximately accurate results. Typically, these are used to increase the speed of a program greatly at the cost of potentially yielding suboptimal results. BLAST and FASTA use heuristics. 

High-throughput DNA sequencing Experimental procedures for determining massive amounts of genomic DNA or cDNA sequence data using highly automated sequencing machines. 

HMM (Hidden Markov Model): The extension of a Markov model. A pattern recognition method that can be used to represent the alignment of multiple sequences or sequence segments by attempting to capture common patterns of residue conservation.HMMER is a software package for profile hidden Markov model analysis. 

Homology: Two genes are said homolog is they derived from a common ancestor. 

Iterative search: After performing an initial search against the database, the high scoring matches are used to search the database again. In some cases (intermediate sequence path), these sequences are used on their own; in others, the sequences are joined together in an alignment or profile. 

Linux: A freely available but commercial-strength clone of the UNIX operating system. A godsend for starting bioinformatics on a budget. It is easily installed alongside Windows on a PC, so the same machine can be booted into either Linux or Windows. 

Local alignment: The alignment of segments from two (or more) nucleic acid or protein sequences. 

Low-complexity region: A region of a nucleic acid or protein sequence with highly biased residue composition, or consisting of many short near-perfect repeats. 

Markov model: A statistical model for sequences in which the probability of each letter depends on the letters that precede it. If the probability depends on the k preceeding letters, the model is said of order k. 

Masking: Some regions of sequences have particular characteristics (such as repeated patterns) that lead to spurious high scores. Masking replaces these "low complexity regions" of sequence with an X (for proteins) or N (for nucleic acids). 

MEDLINE: A free on-line literature database of papers in biomedical sciences (see http://www.ncbi.nlm.nih.gov/Entrez/medline.html). 

Metagenomics: Sequencing and analysis of genetic material retrieved directly from environmental samples. The term metagenome refers to the collection of genes sequenced from metagenomic data. 

Microarray (or DNA chip): An technology that allows to measure simultaneously the expression of thousands of genes in one or several experimental conditions. 

Motif: A short conserved region in a protein sequence. Motifs frequently form a recognition sequence or are highly conserved parts of domains. Motif is sometimes used in a broader sense for all localized homology regions, independent of their size. 

Motif descriptor: A data structure that stores information about a sequence family, motif or domain family. Typical examples are consensus sequences, patterns, profiles and HMMs. 

mRNA expression profile: The identities and absolute or relative expression levels of mRNAs that characterize a particular cell type or physiological, developmental or pathological state. 

Multiple alignment: An alignment of three or more sequences, with gaps (spaces) inserted in the sequences such that residues with common structural positions and/or ancestral residues are aligned in the same column of the multiple alignment. 

Needleman-Wunsch algorithm: The standard dynamic programing algorithm for finding optimal global alignments. 

Neural network: A statistical pattern recognition method. 

Optimal alignment: A global or local alignment of two sequences with the highest possible score. 

Orthologs: Homologous sequences in different species that arose from a common ancestor gene during speciation. Ortholog are often (but not always) responsible for a similar function. See also "paralogs". 

P-value: Probability that an event occurs by chance. 

PAM (Point Accepted Mutation): A collection of substitution matrices, derived by M. Dayhoff, based on phylogenetic reconstruction. 

Paralogs: Homologous sequences (that is, sequences that share a common evolutionary ancestor) that diverged by gene duplication. See also "orthologs". 

Pattern: A descriptor for short sequence motifs, consisting of amino acid characters and meta-characters that can represent ambiguities or variable length insertions. 

PDB database: Protein Data Bank. The repository of solved protein structures. 

Phylogenetic footprinting: A bioinformatic approach to find functional sequences in the genome that relies on detecting their high degrees of conservation across different species. 

Phylogenetic profile: Distribution of homologous genes in species. Comparison of phylogenetic profiles can be used to predict functionnally related genes. 

Position-specific scoring matrix (PSSM): A model representing characteristics of a group of aligned sequences, the simplest form of which is the tabulation of the frequency of amino acids (or nucleotides) in each position of the multiple sequence alignment. Also called "motif profile". 

Proteomics: Technically and conceptually similar to functional genomics, but with the aim of studying biological aspects of all proteins at once in a systematic manner. 

PSI-BLAST: Position-specific Iterated BLAST. An iterative search that uses the BLAST algorithm to provide fast searches, and builds a profile at every iteration. 

Regular expression: A text pattern that conforms to regular grammar and that is used for text pattern matching in the UNIX system, as well as for representing consensus sequence patterns in biology. 

Rooted tree: A phylogenetic tree in which the last common ancestor of all genes, or organisms, displayed on the tree is specified by the initial bifurcation of the tree. 

Score: A number used to assess the biological relevance of a finding. 

Sensitivity: one measure of the performance of a program, define by Sn=TP/(TP+FN) where TP=true positive and FN=false negative. Sn thus measures the proportion of actual positives which are correctly identified as such. A sensitivity of 100% means that the test correctly detects all positive (i.e. does not contain negative in the predictions). See also "specificity". 

Sequence signal: A local functional site in genomic DNA, such as a splice site or a TATA box. 

Smith-Waterman algorithm: The standard dynamic programing algorithm for finding optimal local alignments. 

Specificity: one measure of the performance of a program, define by Sp=TN/(TN+FP) where TN=true negative and FP=false positive. Sp thus measures the proportion of actual negatives which are correctly identified as such. A specificity of 100% means that the test correctly recognizes all negatives (i.e. does not miss positive in the predictions). See also "sensitivty". 

Substitution matrix: The collection of all substitution scores (ex: PAM, BLOSUM). 

Substitution score: The score for aligning a particular pair of letters. 

SWISS-PROT: A curated protein sequence database that provides a high level of annotations, a minimal level of redundancy and a high level of integration with other databases. It is maintained collaboratively by the Department of Medical Biochemistry at the University of Geneva and the European Bioinformatics Institute (EBI). 

Synteny (literally "on the same ribbon"): Co-localization of genetic loci (genes) on the same chromosome within an individual, regardless of whether on not they are phylogenetically linked. Shared synteny (or homology blocks) describes preserved co-localization of genes on chromosomes of related species. Note that conserved synteny does not imply conserved gene order. 

Taxon (pl. taxa): A group of one or more organisms. The term is often applied to the organisms represented by the terminal branches of a tree. 

True positive (often denoted TP): something correctly predicted as positive. Ex: a gene which is correctly predicted to be regulated by a transcription factor X, as it is in reality. See also "false negative" and "false positive". 

Ungapped alignment: An alignment in which gaps are not permitted. 

UNIX: A computer operating system (an alternative to Windows or MacOS). 

Unrooted tree: A phylogenetic tree in which the last common ancestor of all genes, or organisms, on the tree is not specified. 

Z-score: The number of standard deviations from the mean.