Tinderbox User-to-User Forum (for formal tech support please email: info@eastgate.com)
http://www.eastgate.com/Tinderbox/forum//YaBB.cgi
Tinderbox Users >> Tinderbox applications >> Tinderbox and medical literature mining
http://www.eastgate.com/Tinderbox/forum//YaBB.cgi?num=1208898253

Message started by alfredo.polo on Apr 22nd, 2008, 5:03pm

Title: Tinderbox and medical literature mining
Post by alfredo.polo on Apr 22nd, 2008, 5:03pm

Hello all, I am trying to find a way to map and mine medical literature, and be able to find hidden relationships between papers covering distinct topics, matching only laterally. This has been described theoretically and also some workflows have been described using in-house software.

My workflow includes RSS feeding and Bookends for article retrieving from pubmed (just abstract or full text depending on proxy status). DevonThink for collecting text snippets and explore relationships ("see also" feature). Scrivener for drafting and Mellel + Bookends for final writing.

I have been exploring TB for literature mapping. Its nice integration with Bookends is very useful. Link creation and editing is very powerful, and is the feature I am exploiting now. However I am not using either attributes nor agents. In this way, very interesting paths are often discovered in the map (however I am working in Demo mode being very limited) and I imagine this can be further exploited.

I am thus asking if someone has some experience or want to start a project in medical literature mapping, ontology creation and knowledge discovery using TB and PubMed. Suggestions and recommendations will be wellcome.

Thanks

Title: Re: Tinderbox and medical literature mining
Post by Mark Bernstein on Apr 22nd, 2008, 5:51pm

A very fine application -- not limited to medicine, of course!

My advice is not to worry greatly about features you are NOT using; there's no need to use what you don't need.  But, over time, it's nice to know they're there.

I would think that prototypes and Set attributes would both be very useful for working with your ontologies....

Title: Re: Tinderbox and medical literature mining
Post by alfredo.polo on Apr 23rd, 2008, 4:51pm

Of course there are many other applications in the biomedical field: there are published experiences mining PubMed to find protein-protein interactions, gene-disease relationships and others. My field is clinical oncology and I like to find relationships between treatment and outcome,  finding relationships than can trigger new hypotheses. There is some seminal papers [Swanson DR: Medical literature as a potential source of new knowledge. Bulletin of the Medical Library Association 1990, 78(1):29-37] and some other, more recent, revisiting this concept under the new internet paradigm [Bekhuis T: Conceptual biology, hypothesis discovery, and text mining: Swanson's legacy. Biomedical Digital Libraries 2006, 3:2].

The underlying idea is the following (I am quoting Bekhuis 2006): there is "Undiscovered public knowledge", published knowledge effectively buried in disjoint topical domains 'disjoint' because researchers working in disparate fields are unaware of one another. Hence, truly disjoint literatures have no articles in common. Swanson suggested in a series of creative papers that novel information might be unearthed by systematically studying seemingly unrelated and non-interactive research literatures, which he called "complementary but disjoint". To demonstrate the feasibility of his ideas, he found evidence for previously overlooked relationships between fish oil and Raynaud's syndrome, magnesium and migraine, somatomedin C and arginine, and viruses as weapons. This is quite remarkable given that Swanson is a mathematician and an information scientist, not a physician.

The idea of using Tinderbox to map knowledge from PubMed is to create links between single papers, name these links in a very precise way and find new paths (of course you need maybe thousands of PubMed references). Scriptability, to import PubMed-XML references will be desirable (I am bypassing this by dragging references from Bookends, but is slow and too orientated).

These are just a few ideas. I have been testing different software packages, and TB seems ideal for this task. Leopard integration, major scripting capabilities and a few other minor details could be definitely make TB a kill app for biomedical research.

Title: Re: Tinderbox and medical literature mining
Post by Peter Pawelek on May 3rd, 2008, 1:00pm

Hi Alfredo,

I don't want to stray too far off-topic, but can you please post the reference to the use of data mining for protein-protein interaction literature? That interests me greatly.

And, yes, I too am trying to fit Tinderbox into my workflow for literature management/concept visualization/datamining. So far this is a cumbersome beast that involves Tbox, VoodooPad, DevonThink, and Together (previousky KIT). I'm still not satisfied with how it all hangs together. I'd prefer to do everything within Tinderbox, but I haven't quite been able to make that leap.

P


Title: Re: Tinderbox and medical literature mining
Post by alfredo.polo on May 8th, 2008, 5:00pm

Hello, sorry for the delay, I have been out of the office...
There is a number of articles covering datamining relationships gene-protein or protein-protein. However I am not an expert on this topic. You can download a free paper from Dr Marcotte at UCLA in:

http://bioinformatics.oxfordjournals.org/cgi/reprint/17/4/359

On the other hand, I am trying to integrate PubMed with TB via XML. The problem is that although XML export is native in TB, importing XML is not implemented, and is somewhat tricky for a naive like me. However, I strongly think that a more streamlined workflow with XML, or applescriptability can help in the creation of the above mentioned ontologies.

Hope to find more feedback...

Alfredo


Title: Re: Tinderbox and medical literature mining
Post by Mark Anderson on May 8th, 2008, 6:15pm

Outside my area of expertise but as Tinderbox does 'understand' OPML, which is in XML, perhaps XLST could be used to transform the PubMed XML into an XML form that TB can ingest? Over to someone who understands XML/XSLT!

Title: Re: Tinderbox and medical literature mining
Post by alfredo.polo on May 8th, 2008, 6:30pm

I have been doing some experiments... PubMed can export the search to XML. Endnote or Bookends can do online search and export to XML. TB easily understands dropping a XML file onto any view, but it creates a single note with the whole search (perhaps hundreds or thousands of articles) in it. It is needed either a way to manipulate (scripting, actions?) this note body to split the text using the xml tags or a smartXMLimport wizard to put the results of the search into single notes.

I will try XLST as you suggest...

I'll let you know

Alfredo

Title: Re: Tinderbox and medical literature mining
Post by Peter Pawelek on May 8th, 2008, 8:19pm

Hi Alfredo,

Thanks for the link to that Bioinformatics reprint. That's perfect!

Cheers!

Peter


Tinderbox User-to-User Forum (for formal tech support please email: info@eastgate.com) » Powered by YaBB 2.2.1!
YaBB © 2000-2008. All Rights Reserved.