9
$\begingroup$

Recently, there have been a few threads on creating tag synonyms. I think most of these are not really synonyms. The following shows my thoughts about what are synonyms and what are not. I will also talk about meta-tags, as it is related and has been mentioned a few times.

To begin with, I need to point out that concepts are often organized into a certain hierarchy. For example, alignment => nucleotide-alignment => read-alignment. This is related to the discussion below. I would also encourage people to have a look at the tag synonym page at SO. I don't think they are all perfect, but they are very informative.

In the SO tag synonym page, there are real synonyms like "format" vs "formats", "swift-3" vs "swift3" or "amazon-rds" vs "aws-rds". In these examples, the two tags are interchangeable. The relationship is obvious most of time.

What is harder to tell is semi-synonym: "swift3" vs "swift3.1" or "bash" vs "bash-function". For such pairs, one of the tag is conceptually the parent of the other tag. We create such synonyms when the child tag is unnecessarily narrow. Judging "unnecessity" is subjective. We have to take the habit of the community into account when making the decision.

Back to our recent examples:

  • vs . I think they are different enough and are not synonyms. "sam" is also conceptually the parent of "cram", "cSRA", "biohdf" and other binary representations of sam. If most of us still prefer to merge them, I would take "sam" as the master tag.

  • vs . As most others said, they are not because many "mapping" related problems have nothing to do with "read-mapping". Similarly, "read-mapping" is only one type of "alignment". They are not synonyms.

Now meta-tags. I think we might be sometimes confusing a meta-tag with a general tag. For example, "alignment" is a general tag, but it still sufficiently narrows the question down to a specific field. In contrast, the examples on the meta-tag page (e.g. "beginner" and "best-practice") are applicable to almost any fields. Here are some recent discussions related to meta tags:

  • vs . I think the latter is a synonym of the former. In addition to files, I am not sure what other formats we talk about daily. I am also ok to separate them out if others can prove me wrong, but "format" is definitely not a meta tag. A lot of existing questions are unrelated to format.

  • vs . Again, I think the latter is a synonym of the former because in bioinformatics, alignment is almost always applied to sequences or data structures derived from sequences (e.g. profiles, secondary structures or graphs). Alignment has a lot of applications, but it is only a small field in bioinformatics. "alignment" is not a meta-tag, either.

$\endgroup$
  • $\begingroup$ The problem with alignment is that it is used very often both for aligning reads to a genome and for aligning sequences to each other. Yes, both are aligning sequences, but sequence-aligning suggests classic multiple sequence alignments to me while read-alignment makes more sense for NGS analyses. The two are related but very different and each can have very different questions. I think we need different tags for these two types of alignment. $\endgroup$ – terdon Jun 9 '17 at 8:31
  • $\begingroup$ As for format vs file-format it's just a question of clarity. file-format is unambiguous and easy to understand while format is less so. Why not go for the less ambiguous name? $\endgroup$ – terdon Jun 9 '17 at 8:33
  • $\begingroup$ Read-alignment is a type of sequence-alignment. In addition to "file formats", what are the other "formats" we are dealing with? If there are no other formats that are not file formats, the two tags are synonym; if there are others, both tags are necessary. Either way, I don't think we should drop "format". $\endgroup$ – user172818 Jun 9 '17 at 12:07
  • $\begingroup$ Yes, read alignment is indeed a type of sequence alignment. However, the word "alignment" is not sufficient to let you know what you're aligning and the techniques/tools used for read or MSAs are completely different. It is very easy to be an expert in one and not the other (one of the main uses for tags is to help experts find questions to answer). So making the main tags explicit (read-alignment and sequence-alignment for example) instead of the vague (alignment) seems like a good idea. $\endgroup$ – terdon Jun 9 '17 at 12:11
  • $\begingroup$ Same for format. Nothing wrong with it, but file-formats is clearer. I think it makes sense to make format a synonym of file-formats and have the latter as the main tag. The more explicit the tags, the less the chance of their being misunderstood and misused. $\endgroup$ – terdon Jun 9 '17 at 12:13
5
$\begingroup$

In my opinion, the main purpose of tags is to describe content the way, people can follow certain tags.

If we follow this logic, or are not very useful.

In my opinion we should use narrow tags, which describe the problem well enough.

$\endgroup$
  • 1
    $\begingroup$ It is interesting that you see read-alignment and sequence-alignment are distinct topics. To me, reads are sequences and therefore read-alignment is sequence-alignment, i.e. sequence-alignment is the parent of read-alignment. BTW, the title of the first bowtie paper is: "... alignment of short DNA sequences". $\endgroup$ – user172818 Jun 9 '17 at 17:13
  • $\begingroup$ I do not think it is bad to have very general tags (like alignment), that can help to attract attention of competent people to put proper tags on (protein-alignment). However, I agree with non-using mapping tag for its ambiguity (you can map RAD tags or restriction sites to genome, prodein IDs to transcript IDs, ... there are plenty things to map). $\endgroup$ – Kamil S Jaron Jun 9 '17 at 17:50
  • 1
    $\begingroup$ @user172818 conceptually you're correct, but in practice people face very different problems. For sequence-alignment, it will be more about MSAs or pairwise global alignments, for read-alignment, there will be focus on read-mapping tools, I think (although I'm less familiar with the second). In any case, meta-tags are discouraged AFAIU. So we should choose tags, so that people can follow what's interesting conveniently; for that I would prefer having two different tags. $\endgroup$ – Iakov Davydov Jun 9 '17 at 19:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .