1. Optimizing Biological Data Integration
Bioinformatics depends not just on the numbers, but also on correct molecular identification (MI) and mapping between high-throughput platforms.
But the databases and algorithms providing MI disagree wildly.
TCGA (The Cancer Genome Atlas) provides 1000’s of samples and over a dozen platforms for data integration.
Integrating both samples and semantics gives us a way to measure the accuracy of MI for filtering and mapping.
In this way we can evaluate and compare data prep strategies:
ID mapping among genes, transcripts, and proteins.
Algorithms for predicting microRNA targets, for aligning NGS data with reference genomes, for calling copy number variations, etc.
Integrating multiple-platform data correctly will open up a new level of comprehensive systems biology modeling.
We have built some bioconductor R packages to support this work, and published our first application. We look to greatly expand the scope, to aid bioinformaticians and curators.
We pre-process raw data into processed data
ripe for answering medical and biological questions.
We pre-process raw data into processed data
ripe for answering medical and biological questions.
MEANINGFUL Translational Bioinformatics
&
Systems Biology
Good choices
Bad choices
Not so meaningful…
pre-processing choices
(annotation, ID mapping, filtering, algorithms,…)
data for analysis & modeling
raw data
MEANINGFUL Translational Bioinformatics
&
Systems Biology
Good choices
Bad choices
Not so meaningful…
pre-processing choices
(annotation, ID mapping, filtering, algorithms,…)
data for analysis & modeling
raw data