xHMMER3x2: Utilizing HMMER3's speed and HMMER2's glocal alignment mode for large-scale protein domain annotation (Introduction)

Choon-Kong Yap, Birgit Eisenhaber, Frank Eisenhaber, Wing-Cheong Wong

Input Method (FASTA format)   Help & Examples
Paste in textbox or upload a file (maximum 1000 sequences)
(Textbox has a limit of 10,000 characters. Please use file upload instead if this limit is exceeded.)


Mode Selection

A comparison of retention and speed for the different modes used, as well as the proportion of HMMER2 and HMMER3 models (pfam release 29) used.
Mode Sensitivity?   Speed?  H2/H3
domains used
1 - "glocal-mode" HMMER3 95.7% 201x         0 H2 domains
16295 H3 domains
2 - xHMMER3x2 / 500 98.6% 37x     500 H2 domains
15795 H3 domains
3 - xHMMER3x2 / 1000 99.1% 18.5x    1000 H2 domains
15295 H3 domains
4 - xHMMER3x2 / 1500 99.3% 13x    1500 H2 domains
14795 H3 domains
5 - xHMMER3x2 / all 99.8% 7x    3502 H2 domains
12793 H3 domains
6 - glocal-mode HMMER2 100% 1x 16295 H2 domains
        0 H3 domains

Email address
(compulsory for 2 or more sequences input)

Job Title?
(optional, to be used in title of email notification)

Instructions for using the xHMMER3x2 webserver

  1. Input protein sequence(s) in either way:
    • Textbox
      You can copy and paste FASTA sequences into the text box. Do note that there is a limit of 10,000 characters when input through textbox.
    • File upload
      You can upload a file that contains multiple sequences in FASTA format. Currently, we support up to 1000 sequences.
  2. Select a program mode from the drop down list.
  3. Enter an email address (compulsory for multiple sequences, i.e. 2 or more sequences) to receive the http link when the job is completed.
  4. You can enter a job title for your task (optional), to be used in the email notification (only for multiple sequences result sent through email).
  5. Press the Submit Query button to submit your query.

Standalone Program

xHMMER3x2 also comes with a standalone program which allows you to run the program on your local computer. It is written as a perl script and requires both HMMER2's hmmpfam and HMMER3's hmmscan to be pre-installed on your own computer.

To download the source codes, please click here

Please read the README file provided in the tar file for further information on how to run the program.


  1. Go to here for a list of HMMER2 models that are found to be more sensitive than their HMMER3 counterparts at all evaluated FP values.

  2. Click here to download all the swissprot human proteome to Pfam domains annotation, in addition with HMMER2 results (E-value & matching sequence position), for pfam release 29.

  3. Click here to download the 140 unique Pfam domain models that are highlighted for annotation inconsistency as a result of FPR quantification by xHMMER3x2, where they exhibit higher HMMER3 FPR than that of HMMER2 (FPH2-FPH3<-4), though given similar sequence-to-domain alignments.

Feedback Login Site Map
Feedback Login Site Map