Monday, August 30, 2010

NGS viewers reviewed

I gathered up some of the recent free next generation sequence viewers that were capabale of viewing BAM files - and put each through the motions with a few BAM files and reference sequences of various sizes. While there are some great ideas and several choices to be found along the feature spectrum, I think we are still in the dark ages with this stuff. No viewer has really been able to entirely combine usability with performance and analysis capabilities, let alone extensibility and web connectivity.

tview
My take: tview is the barebones, text-rendered viewer that is included with Samtools. People who favor this as their BAM viewer probably think Vim is too polished. Even the very limited navigation is remarkably unituitive (goto a coordinate requires chr:position even if you have just one chromosome, no errors are displayed if you forget chr).
Most resembles which video game: Oregon Trail
Good standout feature: command line access
Bad feature: text display

BamView
My take: Bamview is a wicked fast simple BAM file viewer. It doesn't have much in the way of features, but for cursory examination of BAM files it is more palatable than tview.
Most resembles which video game: Burgertime
Good standout feature: strand split screen
Bad feature: drag selecting a region turns it red, and umm... that's all it does

GenoViewer
My take: The only NGS viewer endorsed by Speak the Hungarian rapper, unfortunately this recent entrant leaves a lot to be desired in terms of performance with large files. GenoViewer is very hard on the eyes - the indiscriminate use of primary colors looks like a kid somehow vomited up the ball pit at Chuck E. Cheese.
Most resembles which video game: Centipede
Good standout feature: promises that "You will not get lost in the details, and can easily figure out the true meaning behind the data. Guaranteed."
Bad feature: graphics

MagicViewer
My take: MagicViewer has come a long way since its initial release last December. With its interesting pie-chart icon renderings of SNP purity, decent treatment of annotation tracks, and improved performance, MagicViewer might soon be a contender to Tablet in the midweight category. The navigation is workable but takes some getting used to - it's unclear when to use scroll bars vs. arrows
Most resembles which video game: Moon Patrol
Good standout feature: primer design tool
Bad feature: some regions are simply not visible - no reference, no reads

Tablet
My take: Hands down the most attractive of the viewers, Tablet is aimed at fostering a delicate balance between performance, features, and aesthetics. Tablet comes with a suite of read views - Packed, Stacked, and Classic - to suit both young children and elderly scientists alike. GFF feature files can be loaded but they appear to merely serve as position indices.
Most resembles which video game: SimCity
Good standout feature: interface
Bad feature: read insertions not displayed correctly



IGV
My take: The Integrative Genomics Viewer is a serious tool for exploring and analyzing large datasets. In addition to viewing, IGV is designed to allow users to extract the kind of hard publishable data that has typically been the domain of Bio* scripts. Like the true product of an Ivy League education, IGV can appear aloof and arrogant to newcomers. The viewer will let you load BAM files and other annotation tracks that have nothing to do with the reference without comment or guidance, then require an extra unitutive click to actually generate a view. While not the easiest or even the best in performance, in terms of generating real queries on your data from the viewer there is nothing comparable.
Most resembles which video game: Dig Dug
Good standout feature: analysis tools
Bad feature: throws a hissy fit if it cannot connect to home server


gbrowse2
My take: gbrowse2 is the AJAXified protege to the venerable generic genome browser. Samtools integration is a recent addition to this highly extensible platform, which has been used for years to display everything from large genomes to small sequencing projects. Of all the viewers here, GB2 provides the best set of visualization tools, such that virtually any biological information that can be rendered linearly has been done so as gbrowse tracks. The lone true web application, gbrowse genomic positions can be hyperlinked or even snapshot-embedded on the web. The web provides the best platform to share visual genomic data among several users. However, a gbrowse2 installation with BAM tracks can be a massive pain to install, configure, and debug ("landmark chrI not found" is the most popular google search in all of bioinformatics). Novices can expect a minefield of historical gotchas and arcane conventions ("Name=" not "name=" field in gff3, bp_seqfeature_load instead of bp_load_gff for gff3, no validator for conf files), and even experienced users are often baffled by cryptic errors that pop up in server logs.
Most resembles which video game: Sissyfight 2000
Good standout feature: hyperlinks
Bad feature: setup

10 comments:

  1. I don't have a dog in the hunt, but regarding your criticism of GBrowse2, the GFF3 spec does state that the correct form is Name= and that it's a reserved term along with the other capitalised attributes:

    http://www.sequenceontology.org/gff3.shtml

    It's not GBrowse's fault you're all doing it wrong ;)

    ReplyDelete
  2. This comment has been removed by the author.

    ReplyDelete
  3. Hello, interesting review. I am representing the team developing the GenoViewer.

    First, I'd like to heavily point out that it's a pre-release version, showing the current state of the project. It is still under development, and far from finished (or perfect, either). Many features will come.

    Second, about the colours: you've got the whole 32 bit colour range to choose. It takes a few clicks in the 'Profiles' menu. And you can create and save numerous individual profiles, so... But we are considering changing the default settings.

    Third, I don't agree it performs poorly with large files.

    At least, I understand fully the resemblance to the Centipede game: it is essential to avoid bugs! :)

    Thanks for reviewing and have a nice day!

    - Miklos Laczik

    ReplyDelete
  4. I'm looking forward to future versions of GenoViewer and the other viewers. As a postscript I'd like to emphasize how much opportunity still exists is in this space, especially with regard to viewers geared toward a niche area like assembly or analysis of structural variation.

    ReplyDelete
  5. What about IGB (Integrated Genome Browser) http://www.bioviz.org/igb/ .

    In my view, it's one of the best ones out there.

    ReplyDelete
  6. My choice: Hawkeye, from the AMOS tools. It's a great tool, very intuitive and with a very nice GUI. May be the bad thing is the performance with large scaffolds (large amount of RAM space required). Great tool.

    ReplyDelete
  7. Thanks for the survey. This is helpful to everyone seeking to discover or develop a better viewer.

    ReplyDelete
  8. UGENE team just released Assembly Browser - feel free to test on your BAMs

    ReplyDelete
  9. Do any of the viewers have a scale on the side? I am using IGV currently, but sometimes hard to see at a glance if there are a lot of reads or just a few.

    ReplyDelete