If you use BART Cancer in your data analysis or publish the results of BART Cancer, please cite the following papers in the main text of your manuscript:

  1. BART Cancer: a web resource for transcriptional regulators in cancer genomes
    Zachary V. Thomas, Zhenjia Wang, Chongzhi Zang
    NAR Cancer, (2021)


BART Results Submission Form

The submission form to get BART results includes two selection menus. The first asks you to select a cancer type providing both common name and acronym. The second asks to select the gene set BART was run on. For each cancer, there are two BART results. One for the upregulated genes in that cancer compared to corresponding normal, and one for the downregulated genes in that cancer compared to corresponding normal. Once these two are selected, click submit to get the BART results.

How to interpret the BART result

The primary BART result is displayed as a table with the following columns:

Other output data files are also available for download.

Transcriptional Regulator Plots

Below the BART Results section, you will find a list of transcriptional regulators and corresponding plots. These display various TR specific data in bubble plot, violin plot, and bar chart form. To view different plots, click the name of the TR from the chart to the left of the plots. Note that likelihood scores for every TR are not available, when selected the plots using that data will display "N.A." or will have altered axes to reflect the lack of data.

Results Process

Patient samples were taken from all cancer types from The Cancer Genome Atlas Program (TCGA). Samples were then clustered based on their expression profiles (log2 of FPKM, with pseudo-count 0.1) using k-means clustering (n=31). The image below shows the results of the clustering and cancer types were re annotated by their cluster.

Clustering Results

Some TCGA cancers had similar expression profiles and were re-clustered as one cancer. Colon adenocarcinoma (COAD) clustered with Rectum adenocarcinoma (READ) to form colorectal adenocarcinoma (COAD_READ). Stomach adenocarcinoma (STAD) clustered with Esophageal carcinoma (ESCA) to form Stomach and Esophageal carcinoma (STES). Additionally, BRCA was separated into two subtypes. The cluster containing 833 samples has been labeled as BRCA_1, for mainly luminal breast cancer (764 in 833 are luminal), and the cluster with 234 samples has been labeled as BRCA_2, for mainly basal breast cancer (176 in 234 are basal). Cancer types that had corresponding normal samples (15 cancer types as indicated by black boxes) were then used to determine differentially expressed genes using DESeq2. Differentially expressed genes were defined as those with adjusted pvalue < 1e-7 and log2(Fold Change) greater than 1 for upregulated genes and less than -1 for downregulated genes. The differentially expressed genes (both up and down) were used as input to BART to predict functional transcriptional regulators in each cancer type.

In order to double-check the identification of differentially expressed genes, Venn diagrams were drawn to compare the number of upregulated genes from the Cistrome Cancer project and the newly identified upregulated genes as seen below (blue is Cistrome Cancer upregulated genes and green is newly identified upregulated genes).