Lee H, Kim SY, Kim DW, Park YS, Hwang JH, Cho S, and Cho JY. SeroHet: Signatures of Interpatient Seroprotein Heterogeneity in Lung, Pancreatic, and Colorectal Cancers. Forthcoming.
SeroHet is a MATLAB-based software to study signatures of interpatient seroprotein heterogeneity in cancer patients.
SeroHet is developed and maintained by Hanjun Lee (MIT/MGH/Broad). SeroHet signatures are also available from here.
SeroHet is implemented in MATLAB. MATLAB version equal to or greater than R2019b is recommended. SeroHet has been tested in Windows, Mac, and Linux environments. Download and decompress the latest stable release of SeroHet to your directory of choice. The current stable release of SeroHet is v.1.2.
Alternatively, if you have git functionality in your system, you can run the following:
git https://github.com/hanjunlee21/SeroHet.git
SeroHet requires a matrix of seroprotein concentrations. The nomenclature and the unit for seroproteins should follow that of the library. Novel seroprotein may follow the nomenclature and unit of your preference. Data from individuals without cancer should be labeled as "Healthy Control".
SeroHet requires three input files:
- a single tab-delimited .txt file containing the concentrations of each seroprotein (#seroproteins-by-#samples)
- a single .txt file containing the names of the seroprotein (#seroproteins-by-1)
- a single .txt file containing the group IDs (#samples-by-1)
Enter a MATLAB session in that specific directory and run the following MATLAB function:
SeroHet('path_matrix','path_seroprotein','path_groupID','outdir');By default, SeroHet normalizes the concentration input based on the original SeroHet cohort. This cohort is based on 584 individuals with or without lung, pancreatic, or colorectal cancers. For normalization, only the seroprotein concentration profile of cancer patients are utilized.
In addition, there are several optional fields in SeroHet that users can utilize. Optional fields in SeroHet are mutually exclusive.
SeroHet(___,Name,Value);If you want to perform statistical analysis on your grouped samples, you can utilize the Statistics functionality of SeroHet. By default, the software will create volcano plots based on Welch's t-test and display the signature-wise Bonferroni-corrected threshold for p-values. To perform this optional statistical analysis, run the following MATLAB function:
SeroHet('path_matrix','path_seroprotein','path_groupID','outdir','Statistics','on');If you have individuals without cancer in your cohort, you can utilize ROC functionality of SeroHet to measure signature-wise ROC curves in your cohort. By default, the software performs 1,000 iterations of 10-fold cross-validation. The aggregate data will be saved as a text file. The software will also create volcano plots for each signature based on the conditional diagnostic odds ratio and its cognate p-values calculated by Student's t-test.
SeroHet('path_matrix','path_seroprotein','path_groupID','outdir','ROC','on');If you want to analyze the relationship between metadata and SeroHet signatures, you can utilize the Metadata functionality of SeroHet. By default, the software will draw a forest plot of signature-wise log2 fold change for each metadata and save a text file containing the -log10 p-values. The input file should be a .txt file containing a #samples-by-#metadata matrix for metadata labels. Currently, only two labels for each metadata column is allowed. For instance, for tobacco smoking-associated column, only "Smoker" and "Non-smoker" labels are allowed.
SeroHet('path_matrix','path_seroprotein','path_groupID','outdir','Metadata','path_metadata');SeroHet produces multiple plots and files as output.
A text file containing SeroHet signature classifications for novel seroproteins
- column 1: seroprotein ID
- column 2: SeroHet signature
A bar plot of logistic(z-score) of each group
- each signature is color-coded and is presented in an order specified here
Figure 1. logistic z-scores of each seroprotein in the initial SeroHet cohort
A volcano plot between two groups
- x-axis: log2 fold change (log2(GroupID2/GroupID1))
- y-axis: -log10 p-value (Welch's t-test)
- each signature is color-coded
- signature-wise Bonferroni-corrected threshold for p-value is presented as a horizontal line
Figure 2. volcano plot showing the alteractions in the expression profiles of each signature between individuals without cancer and AJCC stage I cancer patients in the initial SeroHet cohort
A volcano plot for conditional diagnostic odds ratio in ROC
- only generated when ROC field is turned on
- x-axis: log2 conditional DOR
- y-axis: -log10 p-value (Student's t-test)
- each signature is color-coded
- seroprotein-wise Bonferroni-corrected threshold for p-value is presented as a horizontal line
Figure 3. volcano plot showing the conditional diagnostic odds ratio of seroproteins from SeroHet signature 3 in the initial SeroHet cohort
A text file containing prediction values and state values required for ROC construction
- column 1: prediction values (logistic regression, range: 0–1)
- column 2: state values (0: Healthy Control, 1: Cancer)
A forest plot of signature-wise log2 fold change in each metadata
- x-axis: log2 fold change (log2(MetadataID2/MetadataID1))
- y-axis: metadata type
Figure 4. relationship between SeroHet signatures and metadata in the initial SeroHet cohort
A text file containing -log10 p-values for the signature-wise alterations in each metadata
- column 1: MetadataID1
- column 2: MetadataID2
- columns 3–12: signature-wise -log10 p-value (Welch's t-test)