MACE: Mutation-oriented analysis of chemical response and gene expression in cancer cell lines

What is MACE?
● Integration of drug response data and whole genome gene expression data on cancer cell lines.
● Mutation- and lineage-oriented profiling of chemical screening and omics data on cancer cell lines.
● Speedup of 3D shape similarity search on high-end commodity GPU hardware, as well as 2D-based similarity search.
● Statistical methods to correlate mutation/lineage with drug sensitivity.

Overall diagram of MACE

1. MACE database
● NCI60 drug response data for 50% growth inhibition (GI50) obtained from the NCI Developmental Therapeutics Program (DTP) (http://dtp.nci.nih.gov)
● NCI60 released in 2013 with 50,839 chemicals (75,446 experiments)
● The 2D structure database for 48,616 chemical structures and the 3D conformer database consisting of 2.1 million conformers
● Microarray gene expression data for the NCI60 cancer cell lines obtained from Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) series of GSE32474
● The data preprocessing includes normalization of the primary data, organization of category-oriented profiles, building 2D structure and 3D shape databases for chemicals, and calculation of cross correlation between chemicals, between genes, and between chemicals and genes.
2. MACE web interface
● Typical analysis flow
- Search chemical response profile -> View chemical information -> Select a profile -> View bar graphs -> View a heatmap -> View waterfall plots
- Search gene expression profile -> View cell line-based profile -> Select a profile (mutation-based/lineage-based) -> View bar graphs -> View a heatmap -> View waterfall plots
● Chemical search by chemical name, NSC ID, CAS number, and drawing chemical structure
● Genes search by gene symbol and Affymetrix probe ID
● Three types of profiles for chemicals: normalized individual GI50 data of chemicals against NCI60 cell lines and statistically identified mutation/lineage-specific chemical response in three graphical representations
● Three types of profiles for genes: DNA microarray data on NCI60 cell lines and statistically identified mutation/lineage-specific gene expression in three graphical representations
● Correlation of chemical response profiles and gene expression profiles via similar chemical/gene searches

Mutation-oriented analysis of chemicals and gene
Three types of profiles:

● mutation-specific chemical response/gene expression
● lineage-specific chemical response/gene expression
● normalized log GI50 data of chemicals against NCI60 cell lines/DNA microarray data on NCI60 cell lines

Example profiles of mutation-specific chemical response for NSC ID 741078 in the first row and gene TYR in the second row.

Three different graph types:

● Bar graphs to show the response difference for each of 15 mutation (red bars for significant level < 0.01)
● Hierarchically clustered heatmaps for all profiles
● Waterfall plots for 15 mutations (red dots for cell lines in the specific category)

2D and 3D search results powered by GPU
Examples of 2D and 3D search results
● Structure editor to draw or load a structure
● Selected chemicals in the highest similarity range from the 2D and 3D search results: chemical structures, bar graphs and heatmaps for mutation-based profiles

Combined analysis of chemical response and gene expression data
Examples of profile-based similarity search

● Query chemical: structure, cell line- and mutation-based profiles
● Chemicals and genes with the correlation value more than 0.6 and the significant level less than 0.01
● Similar chemical search result: structure, cell line- and mutation-based profiles
● Similar gene search result:cell line- and mutation-based profiles

Last update: November 2014