BINF-D-17-00305R2

[juanmirocks]

Dear M. Sc. Cejuela,

Your manuscript "LocText: relation extraction of protein localizations to assist database curation" (BINF-D-17-00305R1) has been assessed by our reviewers. Based on these reports, and my own assessment as Editor, I am pleased to inform you that it is potentially acceptable for publication in BMC Bioinformatics, once you have carried out some essential revisions suggested by our reviewers.

Their reports, together with any other comments, are below. Please also take a moment to check our website at http://binf.edmgr.com/ for any additional comments that were saved as attachments.

Once you have made the necessary corrections, please submit a revised manuscript online at:

http://binf.edmgr.com/

If you have forgotten your username or password please use the "Send Login Details" link to get your login information. For security reasons, your password will be reset.

We request that a point-by-point response letter accompanies your revised manuscript. This letter must provide a detailed response to each reviewer/editorial point raised, describing what amendments have been made to the manuscript text and where these can be found (e.g. Methods section, line 12, page 5). If you disagree with any comments raised, please provide a detailed rebuttal to help explain and justify your decision.
Please also ensure that your revised manuscript conforms to the journal style, which can be found in the Instructions for Authors on the journal homepage.

A decision will be made once we have received your revised manuscript, which we expect by 30 Dec 2017.

Please note that you will not be able to add, remove, or change the order of authors once the editor has accepted your manuscript for publication. Any proposed changes to the authorship must be requested during peer-review, and adhere to our criteria for authorship as outlined in BioMed Central's policies. To request a change in authorship, please download the 'Request for change in authorship form' which can be found here - http://www.biomedcentral.com/about/editorialpolicies#authorship. Please note that incomplete forms will be rejected. Your request will be taken into consideration by the editor, and you will be advised whether any changes will be permitted. Please be aware that we may investigate, or ask your institute to investigate, any unauthorized attempts to change authorship or discrepancies in authorship between the submitted and revised versions of your manuscript.

We look forward to receiving your revised manuscript and please do not hesitate to contact us if you have any questions.

Best wishes,

Rezarta Islamaj Dogan, Ph.D.
BMC Bioinformatics
https://bmcbioinformatics.biomedcentral.com/

Technical Comments:

Editor Comments:

Dear authors,

The reviewers have recommended publication, however one of the reviewers has one minor concern which I would like for you to address, and reply back to me. Please send me your comments/response/revision as soon as you can,

Thank you

"
However, I am still concerned about the claims that F-score of LocText is significantly better than Baseline, both for NER and overall. Even the authors admit that both have a borderline support, and as such should receive more careful treatment than just standard error, for instance an explicit Student's t-test. This is especially perplexing while such claims are basically irrelevant to the actual usefulness of LocText (while Baseline method is practically unusable).
The PR-curve plot attached to the comments was particularly illuminating; I regret that authors don't want to append it -- Baseline performance could be shown simply as a point in the P-R space.
"

Reviewer reports:

Reviewer 1: The text has been substantially improved since my last review, and most of my concerns have vanished.

• However, I am still concerned about the claims that F-score of LocText is significantly better than Baseline, both for NER and overall. Even the authors admit that both have a borderline support, and as such should receive more careful treatment than just standard error, for instance an explicit Student's t-test. This is especially perplexing while such claims are basically irrelevant to the actual usefulness of LocText (while Baseline method is practically unusable).
  The PR-curve plot attached to the comments was particularly illuminating; I regret that authors don't want to append it -- Baseline performance could be shown simply as a point in the P-R space.

Regardless, these flaws are minor and I can recommend acceptance in the present form.

Reviewer 2: The authors have addressed the comments (including mine) from the reviewers in the revised version of the MS. I have no further comments and consider this manuscript ready for publication.