Walter M. Fitch

“Homology” in proteins and nucleic acids: A terminology muddle and a way out of it

“Homology” has the precise meaning in biology of “having a common evolutionary origin,” but it also carries the loose meaning of “possessing similarity or being matched.” Its rampant use in the loose sense is clogging the literature on protein and nucleic acid sequence comparisons with muddy writing and, in some cases, muddy thinking In its precise biological meaning, “homology” is a concept of quality. The word asserts a type of relationship between two or more things. Thus, amino acid or nucleotide sequences are either homologous or they are not. They cannot exhibit a particular “level of homology” or “percent homology.” Instead, two sequences possess a certain level of similarity. Similarity is thus a quantitative property. Homologous proteins or nucleic acid segments can range from highly similar to not recognizably similar (where similarity has disappeared through divergent evolution). If using “homology” loosely did not interfere with our thinking about evolutionary relationships, the way in which we use the term would be a rather unimportant semantic issue. The fact is, however, that loose usage in sequence comparison papers often makes it difficult to know the authors intent and can lead to confusion for the reader (and even for the author). There are three common situations in which hazards arise by using “homology” to mean similarity. The first case is the most obvious offense but perhaps the least troublesome. Here an author identifies sequence similarities (calling them homologies) but claims that the sequences being compared are not evolutionarily related. Some awkward moments occur in such a paper, since the author claims both homology (i.e., similarity) and nonhomology (i.e., lack of a common ancestor). Nonetheless, the author’s ideas are likely to be clear since arguments against common ancestry are presented explicitly. A second case is one in which an author points out similarities (again called homologies) but does not address the issue of evolutionary origins. The reader, seeing the term “homology,” may infer that the author is postulating coancestry when that is not the authors intent. The final case occurs most frequently and is the most subtle and therefore most troublesome. Here, similarities (called homologies) are used to support a hypothesis of evolutionary homology. In this case, the two meanings of homology seem to overlap, and it is almost inevitable that the thinking of author and reader alike will be intrusively distorted as follows. Similarity is relatively straightforward to document. In comparing sequences, a similarity can take the form of a numerical score (O/o amino acid or nucleotide positional identity, in the simplest approach) or of a probability associated with such a score. In comparisons of three-dimensional structures, a typical numerical description is root-mean-square positional deviation between compared atomic positions. A similarity, then, can become a fully documented, simple fact. On the other hand, a common evolutionary origin must usually remain a hypothesis, supported by a set of arguments that might include sequence or three-dimensional similarity. Not all similarity connotes homology but that can be easily overlooked if similarities are called homologies. Thus, in this third case, we can deceive ourselves into thinking we have proved something substantial (evolutionary homology) when, in actuality, we have merely established a simple fact (a similarity, mislabeled as homology). Homology among similar structures is a hypothesis that may be correct or mistaken, but a similarity itself is a fact, however it is interpreted. We believe that the concepts of evolutionary homology and sequence or three-dimensional similarity can be kept distinct only if they are referred to with different words. We therefore offer the following recommendations: *Sequence similarities (or other types of similarity) should simply be called similarities. They should be documented by appropriate statistical analysis. In writing about sequence similarities the following sorts of terms might be used: a level or degree of similarity; an alignment with optimized similarity; the % positional identity in an alignment; the probability associated with an alignment. *Homology should mean “possessing a common evolutionary origin” and in the vast majority of reports should have no other meaning. Evidence for evolutionary homology should be explicitly laid out, making it clear that the proposed relationship is based on the level of observed similarity, the statistical significance of the similarity, and possibly other lines of reasoning. One could argue that the meaning of the term “homology” is itself evolving. But if that evolution is toward vagueness and if it results in making our scientific discourse unclear, surely we should intervene. With a collective decision to mend our ways, proper usage would soon become fashionable and therefore easy. We believe that we and our scientific heirs would benefit significantly.

Influenza B virus evolution: Co-circulating lineages and comparison of evolutionary pattern with those of influenza A and C viruses

Sequence analyses and comparison of the genes coding for the nonstructural (NS) and hemagglutinin (HA) proteins of different influenza B viruses isolated between 1940 and 1987 reveal that the number of substitutions is not always proportional to the time between isolates. Examination of 14 influenza B virus NS gene and 10 HA gene sequences by the maximum parsimony method suggested that--as with influenza C viruses--there are multiple evolutionary lineages which can coexist for considerable periods of time. Comparison of the sequence divergence among genes of viruses belonging to type A, B, and C virus suggests that, in man, influenza B viruses evolve slower than A viruses and faster than C viruses. We propose an evolutionary model for influenza B viruses that is intermediate between the pattern for human influenza A viruses and that for influenza C viruses.

Finding the Minimal Change in a Given Tree

It is a common task in biology to determine the genealogy of species, populations, people, or genes and estimate the condition of the ancestral forms. That is often done for molecules such as proteins and nucleic acids. There are many procedures. I address here only the parsimony procedures which ask, “ How can I account for the descent of these various sequences from a common ancestor with the fewest number of changes?” The general problem being addressed is as follows. One has a set of s sequences, each sequence being a linear string of letters from some alphabet. In molecular biology the sequences are either proteins, of which there are 20 letters (amino acids), or nucleic acids, of which there are 4 letters (nucleotides). We assume that the sequences have been aligned by some method so that they are all of the same length, t. It is assumed that these s sequences arose from a common ancestral sequence by a branching process that is properly described as a strictly bifurcating tree, that is, as a graph in which there is one and only one path connecting any two nodes on the tree. The tree has s tips (exterior nodes of degree one), one for each of the s sequences and s — 2 interior nodes of degree three, plus one node of degree two, called the root, that is the ultimate ancestor, the node at which the branching process began. The edges connecting two adjacent nodes are called branches. The task is to discover for any given tree topology, the minimum amount of change, and its nature, on each branch.

Glycoprotein evolution of vesicular stomatitis virus New Jersey

A T1 ribonuclease fingerprinting study of a large number of virus isolates had previously demonstrated that considerable genetic variability existed among natural isolates of the vesicular stomatitis virus (VSV) New Jersey (NJ) serotype [S.T. Nichol (1988) J. Virol. 62, 572-579]. Based on these results, 34 virus isolates were chosen as representing the extent of genetic diversity within the VSV NJ serotype. We report the entire glycoprotein (G) gene nucleotide sequence and the deduced amino acid sequence for each of these viruses. Up to 19.8% G gene sequence differences could be seen among NJ serotype isolates. Analysis of the distribution of nucleotide substitutions relative to nucleotide codon position revealed that third position changes were distributed randomly throughout the gene. Third base changes constituted 84% of the observed nucleotide substitutions and affected 89% of the third base positions located in the G gene. Only three short oligonucleotide stretches of complete sequence conservation were observed. The remaining nucleotide changes located in the first and second positions were not distributed randomly, indicating that most of the amino acids coded by the G gene cannot be altered without reducing the fitness of the VSV NJ serotype viruses. Despite these constraints, up to 8.5% amino acid differences were observed between virus isolates. These differences were located throughout the G protein including regions adjacent to defined major antibody neutralization epitopes. Apparent clusters of amino acid substitutions were present in the hydrophobic signal sequence, transmembrane domain, and within the cytoplasmic domain of the G protein. A maximum parsimony analysis of the G gene nucleotide sequences allowed construction of a phylogram indicating the evolutionary relationship of these viruses. The VSV NJ serotype appears to contain at least three distinct lineages or subtypes. All recent virus isolates from the United States and Mexico are within subtype I and appear to have evolved from an ancestor more closely related to the Hazelhurst historic strain than other older strains. The implications of these findings for the evolution, epizootiology, and classification of these viruses are discussed.