BRAVO: Biomarker Recognition And Validation Online

We describe a new data mining system for Biomarker Recognition And Validation Online (BRAVO) program. BRAVO consists of two integrated components: a large multi-disease multi-cohort patient database and a gene-centric statistical data mining engine which receives users online query, performs statistical analysis and packages results back to the user. BRAVO provides users with easy access to thousands of patient samples and analysis tools for many important biological questions with similar diseases. One advantage of BRAVO is that it automatically finds patients cohorts that match user’s query and performs the analysis on multiple cohorts in parallel. This feature significantly reduced the user’s labor comparing with using other tools such as Oncomine because BRAVO allows the users to focus on their own questions rather than the target cohorts. Another feature of BRAVO is that all collected public data have gone through a rigorous examination and normalization process by our experts to ensure data integrity. The genomics information generated from BRAVO will provide better understanding of biomarkers development toward personalized medicine for cancer treatment and diagnosis

Who may be benefited using BRAVO?

You have one or more of candidate biomarkers (genes) of interests You want to validate the gene expression in published patient samples You want to test whether there are signficant association between a pair of genes or between a gene and other clinical factors in the patient cohorts

Which datasets does BRAVO contain?

The cancer patient database includes a large and high quality selected collection of publicly available patient datasets containing both gene expression and clinical data such as disease stage/grade, patient survival time/status, histological and molecular subtypes. Currently, BRAVO contains data for Breast, Colorectal, Prostate, Lung, Ovarian, and Brain cancer with more than 30 cohorts and 5 thousand patient samples.

How to use BRAVO?

Step1: select the query type

Biological queries were formulated into 5 categories: co-expression of a pair of genes; differential expression between 2 groups; gene expression vs. continuous variable; gene expression vs. discrete variable; gene expression vs. survival correlation.

Step2: select disease

A popup window will open to show you all the clinical phenotypes for that disease. You will need to use these variable names to select the subset of samples of interests. You also need to use the variable names if you want to calculate correlations between a gene and that variable. Note that you need to use the EXACT variable name.

Step3: input gene symbol

Step4: select samples (for some query type)

The default (empty field) will select all samples. Otherwise you need to input an logical expression to select samples. See examples below

Step5: input phenotype or survival(for some query type)

What are continuous and discrete phenotypes?

Continuous phenotype examples: AGE_AT_DIAGNOSIS_YEAR, TUMOR_SIZE_MM

Discrete phenotype examples: GENDER, ER_STATUS

How to write a group selection expression?

You can use relational operators: = (equal), >(greater than), < (less than), >= (greater than or equal to), <= (less than or equal to), !=(not equal)

You can use logical operators: and, or

Your can use parathesis: (, )

Be carefuly with logical operators. If you want low grade tumor samples grade 1 and grade 2. You may write:

grade=1 or grade=2

The expression

grade=1 and grade=2

will fail to select any samples. See examples below for detail.

Why do I have no results?

Either there are no samples that satisfy your sample selection criteria, or you have a typo in the gene symbol or phenotype name

Examples

differentially expression of STAT3 between grade 1 and grade 3 patients

query type: differential expression
disease: breast cancer
group1: GRADE = 3
group2: GRADE != 3
gene: STAT3
		
		

co-expression of TWIST1 and CCL5 in triple negative breast cancer

query type: co-expression
disease: breast cancer
group: ER_STATUS = negative and PR_STATUS = negative and ERBB2_HER2_STATUS = negative
gene1: TWIST1
gene2: CCL5
		

		

ADD3 expression correlation to disease free survival in colorectal cancer

query type: survival
disease: colorectal cancer
gene:  ADD3
survival: DFS
		

		

correlation between STAT3 expression and percentage of tumor in breast cancer

query type: gene-continuous
disease: breast cancer
gene: STAT3
pheno: PERCENT_TUMOR
		

		

correlation between STAT3 expression and GRADE in male lung cancer patients

query type: gene-discrete
disease: lung cancer
gene: STAT3
pheno: GRADE
group: GENDER = male

		

all correlations between STAT and BRCA genes in high grade breast cancer patients

query type: gene-gene-list
disease: breast cancer
gene1: STAT3 STAT4
gene2: BRCA1 BRCA2
group: GRADE = 3 or GRADE = 4
p-value type: none
		
		

How do I cite BRAVO?

Deng X, Warden C, Liu Z, Zhang Ian and Yuan YC, BRAVO: biomarkers recognition and validation online, in submission

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes 1: (example: STAT3)
Genes 2: (example: STAT3)

Sample Conditions
Group: (example: grade = 3 , see instructions)

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes 1: (example: STAT3 STAT5)
Genes 2: (example: STAT3 STAT5)

Sample Conditions
Group: (example: grade = 3 , see instructions)

P-Value Cutoff
P-Value < (example: 0.05 , see instructions)

Select P-Value Type
False-Discovery Rate (FDR)
Unadjusted P-Value
None

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes: (example: STAT3)
Sample Conditions
Group 1: (example: grade = 3, see instructions)
Group 2: (example: grade = 3, see instructions)

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Sample Conditions
Group 1: (example: grade = 3, see instructions)
Group 2: (example: grade = 3, see instructions)

Fold-Change Cutoff
|Fold-Change| > (example: 1.5 , see instructions)
P-Value Cutoff
P-Value < (example: 0.05 , see instructions)

Select P-Value Type
False-Discovery Rate (FDR)
Unadjusted P-Value
None


Select Output Type
One File with All Differentially Expressed Genes
Separate Files for Up- and Down-Regulated Genes

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes: (example: STAT3)
Sample Conditions
Group: (example: grade = 3, see instructions)
Phenotype

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes: (example: STAT3)
Sample Conditions
Group: (example: grade = 3, see instructions)
Phenotype

Select Disease
(A window will open to display the data sets and varaiables that you can use)

Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Select Survival Type
Overall Survival Disease Free Survival

Gene Symbols
Genes: (example: STAT3)
Sample Conditions
Group: (example: grade = 3, see instructions)

Select Disease
(A window will open to display the data sets and varaiables that you can use)



Breast Cancer	Brain Cancer	Colorectal Cancer
Leukemia	Lung Cancer	Ovarian Cancer
Prostate Cancer

Gene Symbols
Genes: (example: STAT3)
Clinical Variables
Metadata: (example: TUMOR_STAGE ER_STATUS, see instructions)

The following persons have contributed to the development of BRAVO:

• Xutao Deng: project design, query engine and interface programming
• Charles Warden: query engine and interface programming, data processing
• Zheng Liu: query engine programming, data processing
• Ian Zhang: data processing
• Yate-Ching Yuan: project design and coordination

• BIC wiki
• City of Hope