Supporting Reasoning within Use Case WP7b

July 12th, 2011, 12:00 - 12:50 CET

Participants: Zhisheng (VUA), Mark (UFSD), Niraj (UFSD),

Agenda

Minutes

 gwas:x  gwas:searchInRif "false"

 true: use RIF data,  false: use PubMed data

 gwas:x  gwas:useUMLS "false" .

true: use UMLS keyword extension, false: use UMLS keyword extension

 gwas:x  gwas:searchMode "1" .

1: all data without considering the date constraint,  2: before the date, 3: after the date

 gwas:x  gwas:dateConstraint "20110412"

what does snpHasScore mean? Is it prior propability? or BFDP (Baysian false-discovery probability)?

answer: it is prior probability.

If the answer is for the former, how can we get BFDP of SNPs?

answer: BFDP requires two things. prior and additional calculation

Does the keyword group change the results or not?

gwas:x gwas:hasKeywordGroup gwas:g1

Answer: newly found keywords by the keyword reasoner should be marked as the same group.

how can we organize the experiments next?

Three diseases:

Breast cancer keywords

Prostate cancer keywords

Type 2 diabetes keywords

Try on all the combinations:

1) MeSH or LLD

2) Semantic Distance Measure or not

3) Google distances or others.

4) Substring match or not

5) Direct subClass or not

6) Candidate SNPs, if without any SNP in the queries, it would consider all the SNPs in the file MIMIX/mimix-indices/snp-index/indexedSNPs.txt

It may take a week to get the results.

splitting the SNP file into multiple subfiles, and test it with the same file name of all SNPs list file to test the whole set separately.

What are the criteria for the evaluation?

answer:

1) prior

2) time performance

is the existing MIMIR data enough for the experiments?

answer: got everything Sheffield has.

is it sufficient to get meaningful results?

answer: if it is possible to add more from ?PubMed.

August 2nd, 2011 12:00 - 13:00 CET

WP4/Meeting&TeleConf/Minutes/142July2011 (last edited 2011-07-12 10:57:22 by ZhishengHuang)