PathBinderH Introduction

PathBinder is a mid-way database for enhanced access to PubMed information. Enhancement includes:

• more relevant information retrieval (sentences containing atleast 2 terms of interest)

• ontology-based information retrieval (terms are from several ontologies)

• taxonomy filter

In other words, with PathBinder, a user can retrieve PubMed information using such queries: show me all the sentences containing both "aminobutyraldehyde dehydrogenase" and "central root cap," and I am only interested in the sentences from the abstracts which talk about "maize" or "rice."

In this prototype version of PathBinder, the dictionary includes ~20,000 terms (~33,000 with synonyms) from three ontologies: Enzyme nomenclature, Gene Ontology and Plant Ontology. A subset of PubMed (~65,000 abstracts) was searched for sentences containing at least 2 terms.

How to use PathBinder

Use the following link to view a step by step procedure with a sample query and relavant screen shots

Using PathBinderH: a Sample Query One Step at a Time

In general following are the main steps in the process

1. Setup taxonomy filters (optional)
   Find the taxonomy names of your interest by browsing the taxonomy tree. Add them to your filter list by just clicking the [] icon. For this prototype this feature is available only for the green plants family(NCBI taxonomy ->cellular organisms ->Eukaryota ->Viridiplantae). The sample taxonomy tree has selected taxonomy terms. There is also a list of popular species list from which you could set filter.
   

2. Specify term1
   Find term1 either by drilling down in the ontology trees or through index. The ontology tree contains all the terms and the terms marked "hit!" are in the PathBinder term hits. Clicking on this link will display the list of candidate term 2 which co-occur with term1 at least in one sentence. The number of candidate term 2 is mentioned after the "hit!" sign.
   

3. Specify term2
   Clicking on one of the term from term 2 list will display all the sentences containing both term 1 and term 2 (including their synonyms).
   

4. Display sentences and link to PubMed
   All sentences containing both term1 and term2 (including their synonyms) will be displayed after the two terms are specified. The terms are highlighted. The PMID in the first column is a link to the original PubMed abstract.

Future work

• Scanning against the entire Medline;

• Including (a subset of) UMLS MetaThesaurus in the dictionary;

• More intuitive, user-friendly interface;

• Providing keyword search function;

• Supporting multiple-term1 retrieval;

• Supporting "include-children" retrieval;
  If term1 is a non-leaf term in one of the ontologies, a user may have the option whether to include all its child terms as term1.

• Providing sentence ranking systems;
  Candidate ranking systems: statistical properties, verbs, link grammar parsing...

• Synchronize the PathBinder database with PubMed, NCBI taxonomy, and Ontology source databases, periodically.

• Your requests.
  Your input and feedback are highly appreciated (contact us).