1
$\begingroup$

I am a university student at the department of Biology. As part of my assignment for the subject Bioinformatics, I was asked to refer to common mistakes in biological database (e.g. Uniprot). For instance, "dehydrofolate reductase". Is there any list with such common mistakes? Otherwise, in your own experience, could you mention some examples of similar mistakes that you meet regularly?

$\endgroup$
  • $\begingroup$ "dehydrofolate reductase" does not meet my definition of common. It is used in 14 out of 71 000 000 trEMBL entries in Uniprot. trEMBL is comprised of computationally annotated entries which are explicitly awaiting curation. Is there an error? Dehydrofolate reductase type I (Q4QZB8) has a meaningless name, but plenty of proteins or genes suffer from that fate. The gene ontologies and domain annotations correctly indicate the presence of dihydrofolate reductase. $\endgroup$ – Michael_A Dec 21 '16 at 7:13
  • 4
    $\begingroup$ If there were such lists (maintained by the databases) then why would the errors be still present? This is a manual curation assignment. You have to go through them one by one. $\endgroup$ – WYSIWYG Dec 21 '16 at 8:18
6
$\begingroup$

Microsoft Excel has a nasty habit of thinking it knows your data better than you do, and when it sees gene names like Mar3 and Sept7, it tries to be helpful and convert them to dates.

So anytime someone puts a long gene list in to Excel and takes it back out again, you see those genes and their relatives converted to dates.

This has been known for years, but you still see it in lots and lots of published papers.

According to http://dx.doi.org/10.1186/s13059-016-1044-7, almost 20% of article supplementary material gene lists made with Excel contain such errors.

$\endgroup$
1
$\begingroup$

Sequencing artifacts in the type A influenza databases and attempts to correct them. is a paper that describes errors in one biological database. There are several other papers that similarly address errors in influenza genome databases, but I particularly liked this one because of its unique approach:

As part of a high school class project, influenza sequences with possible errors were identified in the public databases based on the size of the gene being longer than expected, with the hypothesis that these sequences would have an error. Students contacted sequence submitters alerting them of the possible sequence issue(s) and requested they the suspect sequence(s) be correct as appropriate.

Kudos to the lead investigator, David Suarez, for combining a genuinely useful project with cool outreach.

$\endgroup$
  • $\begingroup$ Interesting though this is, it isn't the kind of textual mistake that the poster is referring to. $\endgroup$ – David Dec 21 '16 at 17:14
  • $\begingroup$ "could you mention some examples of similar mistakes that you meet regularly" - I think the paper covers this part of his question pretty well $\endgroup$ – iayork Dec 21 '16 at 17:22
1
$\begingroup$

Strictly speaking I should just vote to close this question as homework or too broad, or repeat WYSWYG’s comment as an answer. However if one changes the question to how one should go about this problem as a bioinformatician, it becomes more interesting.

For a Biology student with no programming skills, it seems something of a distortion to call it bioinformatics. One aspect of bioinformatics is to use programming techniques that automate repetition and if/else decision-making to avoid laborious manual operations. So can one think of possible non-programming approaches? Here are a couple of suggestions to which others might wish to add.

  1. Make up a list of terms like nitrate, phosphate, citrate, sulphate/sulfate, which all end in -ate. In German these are generally similar but end in -at (without the e). Even if you have no German you can use an on-line translator. In my experience some entries in the PDB have mis-spellings arising from such linguistic similarities, so you could search for the German spellings. You might even search for German or French abbreviations for RNA and DNA (RNS, DNS, ARN, ADN). Still a manual approach, but it would have some logic behind it.

  2. If it is possible to find all the key words in your database of interest, can you find how many times each occurs. I could do this with a program, but perhaps Excel has some facility for doing this. You could Google to see. Then you could make the assumption that mis-spellings would only occur a relatively small number of times. You could then work through the terms with the lowest frequency of occurrence, looking for mis-spellings.

If you had some programming ability you could modify approach 2 by looking for words that differ by a single letter, on the assumption that some of these differences would be spelling mistakes. (Presumably a language such as Perl that is designed for use with regular expressions would be most appropriate.)

$\endgroup$
1
$\begingroup$

If you find any such errors in curated databases such as UniProt, please let the database curators know.

In the expert-curated, reviewed section UniProtKB/Swiss-Prot, we'll be happy to fix typos and other errors. As far as the unreviewed UniProtKB/TrEMBL records are concerned, this is slightly more complex as Michael_A states above. However, in many cases, the nucleotide sequence databases (EMBL-Bank/GenBank/DDBJ) can be asked to correct the underlying source, and the corrected names can then be updated in/reimported into UniProtKB/TrEMBL.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.