Levels of Database Match

1.  Specific function: indicated by a specific name and gene symbol
   
   • The protein translation has a good database match to a protein that whose function and process have been experimentally characterized. Both pairwise and multiple protein sequence alignments reveal a high degree of identity/similarity (typically >35% identity) along the entire length of the protein.  There may be an essentially full-length match to a highly specific (i.e., 'equivalog' isology type) HMM.  Active sites, substrate or cofactor binding sites, or motifs that are characteristic of a protein should be conserved. Strong conservation of gene context (e.g. operon structure) is also taken as evidence for certain function. For genes with a certain function we use the most widely-recognized name and gene symbol. Highly specific GO function and process terms are used if available. Enzymes of certain function are annotated with their full IUBMB number; the IUBMB enzyme name may also be used for clarity if it seems more informative than its common name.
     

2. Likely (or unlikely) function: indicated by "putative" or "homolog" in the name
   
   • If one or more lines of evidence is weak, but most of the data agrees, we conclude the gene is likely performing the function the name implies, and the name is preceded with "putative". In such cases, the percent identity (e.g., 30-35% identity) or HMM score (score is between the trusted and noise cutoffs) is not quite high enough to impart certainty. GO terms may be more general than for gene models of certain function, and with few exceptions, gene symbols are not used. For an enzyme with a putative specific function partial IUBMB numbers may be used.
     
     
   • When there are strong lines of conflicting evidence, we consider the function indicated by the common name to be unlikely, and add 'homolog' to the common name. Such assignment can arise from two situations. In the first situation, sequence homology is very strong, but unlike a ‘putative’ match, we do NOT believe the query protein has the same function as the match. This might be because some critical piece of evidence is absent (e.g., non-conservation of catalytic residues in an enzyme), or because the function is not predicted to exist in this particular organism (e.g., photosynthetic enzyme matches in a non-photosynthetic organism).  In the second situation,  there is an essentially full-length match to a set of genes whose names are the same or similar, at least one of which has some experimental characterization, but because the sequence conservation (e.g., 25-30% identity) falls below even the 'putative' range, we consider functional conservation to be unlikely. Furthermore, there are no family or domain names available.  In this case we use the matching proteins' name but add 'homolog' to it, and apply descriptors appropriate for a protein of unknown function.
     
     
   • Note that while using 'homolog' to denote non-conserved function in high-quality matches has been a long-time practice of TIGR annotators, using 'homolog' to retain the names of lower-quality matches that might otherwise be called 'conserved hypothetical proteins' is a relatively recent practice. Also, the criteria for 'putative' annotation has been made more rigorous.  Therefore, it is likely that some older gene models that were called 'putative' or 'conserved hypothetical" would be called 'homolog' by the newer naming criteria.
     

3. Generic function: indicated by protein family name or domain name.
   
   • When the best (or only) annotation evidence indicates membership in a defined family, but does not justify more specific naming, we use family names defined by a TIGR or Pfam HMM(s), curated databases such as SwissProt, or in the literature, e.g., "carbohydrate kinase, FGGY family".
   
   
   • When the extent of sequence homology is limited to a defined protein domain (usually modelled as an HMM), rather than a defined family or full-length characterized protein, we may use the domain name, e.g., "ABC1 domain protein". Since domains are themselves often used to define a family in the literature, the distinction between family and domain based names is not rigid.
   
   
   • Note that the cellular function or process associated with a protein family or domain may be experimentally defined to some degree; alternately, they may be functionally uncharactized, in which case the family or domain name connotes nothing more than sequence-based homology.  The degree of functional characteriziation of the family or domain will be reflected in the degree of specificity of the role categories, enzyme number, and GO term descriptors assigned to the gene model. Thus, when the family or domain has no defined function, the family/domain name are used but the GO terms and role category are the same as for a protein of unknown function (see below).
     

4. Unknown function: conserved hypothetical proteins, conserved domain proteins, proteins of unknown function, and hypothetical proteins
   
   • When the best evidence associated with a protein translation consists only of full-length matches to conceptual translations in other species -- i.e., there are no experimentally characterized matches above ~25% identity, no trustworthy HMM matches, no family or domain or gene context based names that can be reasonably derived from the evidence -- the gene model is annotated as a "conserved hypothetical protein". These are assigned the most general GO terms.   (An exception to this naming practice is made when there is a match to a lipoprotein motif or detection of substantial hydrophobic regions; these are called 'putative lipoprotein' and 'putative membrane protein', respectively.) When the conditions for "conserved hypothetical" apply, but the conserved region is considerably less than the full length of the matches, the gene model is named a "conserved domain protein" instead, but again given the most general GO terms. Note that when matches are only to genes from different strains of the same species, the gene model is considered a "hypothetical protein" (see below).
     
   
   
   • Where good matches exist to proteins that are likely to be real and have been assigned a gene symbol, but for which there are no clues about function -- e.g., the match has been sequenced and shown to be translated in vitro or in vivo, and may even be part of an operon, but no function or genetic effect has yet been determined for it -- the gene model is considered likely to be a real protein of unknown function. We assign the current common name and gene symbol to our gene model, but still use the most general GO terms. As indicated above under "Generic function", proteins of unknown function may have functionally uninformative HMMs associated with them, which may also serve as a basis for names, gene symbols, and general GO terms.
   
   
   • When a gene model is identified by the gene-finding algorithm but has no significant sequence similarity to any characterized or uncharacterized genes from another species or to any defined HMMs, nor an informative gene context, we consider it dubious and name it a "hypothetical protein". Such gene models do not get GO terms or any other descriptors.
     

Alternative gene structures (functional)

1. programmed frameshifts
   
   • When a gene model contains an in-frame termination codon and a naturally-occurring frameshift prior to the termination codon regulates translation of the gene model, we add "programmed frameshift" to the common name.
     

2. selenocysteine-containing proteins
   
   • In certain organisms the 'stop' codon TGA encodes the amino acid selenocysteine. The genome must contain a selenocysteine-tRNA and the enzyme selenide, water dikinase. Proteins which meet these criteria have 'selenocyteine-containing' added to their common names.
     

3. intein- or intron-containing proteins
   
   • When a gene model contains an intein -- a segment of a protein that is able to excise itself and rejoin the remaining portions (the exteins) with a peptide bond -- we add "intein-containing" to the common name. Inteins are also known as "protein introns".
     

Disrupted genes (nonfunctional)

1. authentic frameshifts/authentic point mutations
   
   • When a gene model is disrupted by a either a single frameshift or point mutation, confirmed by manual review of the genome assembly, we simply add "authentic frameshift" or "authentic point mutation" to the common name.
     

2. multiple/mixed frameshifts and point mutations: indicated by 'degenerate'
   
   • When a gene model is disrupted by multiple frameshifts or a mixture of frameshifts and point mutations we assume that the gene model is not functionally expressed and we denote this with the term "degenerate" after the common name.
     

3. interruptions
   
   • Interruptions are cases in which conserved N- and C-terminal portions of a gene model are separated by some other untranslated sequence, such as a transposon. Such gene models are split into two parts with the same common name, with "interruption-N" and "interruption-C" added for the part N-terminal and C-terminal to the interruption, respectively.
     

4. truncations, fragments, and internal deletions
   
   • When a significant segment of the gene model is missing from the N- or C-terminal end -- enough so that we believe that it is no longer functionally expressed -- we add 'truncation' to the common name.
   
   
   • In contrast to truncations, where either the N- or C-terminal region is present, fragments are coterminous only with an internal region of a conserved gene; they lack both the N- and the C-terminal regions of the match. For such gene models we add 'fragment' to the common name.
   
   
   • An internal deletion is the opposite of a fragment: the N- and C-terminal regions are conserved, but the region between them has been deleted. Internal deletions tend to be shorter than interruptions, but are long enough such that we expect the deletion to impair function.  We denote them by adding 'internal deletion' to the common name.
     

5. fusions
   
   • Two proteins which have been fused into one reading frame by an event which deleted a C-terminal portion of protein 1 and an N-terminal portion of protein 2 are denoted by 'fusion' in the common name.