clinco

Work In Progress.

Summarize the clinical (or lab) components of your dataset. Get an overview of the associations and correlations in your dataset.

What is clinco?

clinco is designed to give you a comprehensive view of your dataset so that you can design your analyses. It let's you address questions like:

1. Does my data have batch effects?
2. Is age correlated with expression (or methylation or ...)?
3. Does atopy vary by asthma status (correlations within clinical variables)
4. Which clinical variables are associated with the first X principal components?

Usage

Everything is dispatched from the single command-line executable:

clinco

Data Format

=== clinical data format ===

Clinco expects that your clinical-information is in tab-delimited format with a header indicating the clinical variable in each column one row for each sample. The first column must be some unique ID if the clinical data is to be join to some numerical data such as expression or methylation.

An example with 3 samples would be:

#individual_id	age	sex	date_processed	location
individual_01	22	M	14-01-2012	Denver
individual_02	29	F	22-11-2011	Boston
individual_03	20	F	12-03-2012	Denver

=== numerical data format ===

The numerical data is expected to have n_samples * (n_features + 1). Where each feature is a probe or a measurement. The first column is expected to be an ID that maps that sample back to the clinical data. An example matching the above clinical data is:

#individual_id	probe1	probe2	probe3	probe4
individual_01	0.11	0.91	0.93	0.14
individual_01	0.01	0.71	0.72	0.01
individual_01	0.14	0.99	0.99	0.16

Example

Using data provided in this repo, the command:

python clinco/pca.py \
    -X data/xs.txt \
    -c data/clinical.txt \
    -k gender \
    -f data/ex.pca.png

Does pca on the data in xs.txt with clinical data described in clinical.txt. It will save a plot of the projected data to data/ex.pca.png and print out the correlations of any of the clinical data columns to the first 10 principal components. The plot will vary the colors by gender.

The text output looks like this:

component    clinical_var   n   R   anova_groups    p_value
1            gender         99  na  2-groups        1.7e-55
3            asthmatic      99  na  2-groups        0.0838

Showing that we have a very strong separation of gender on the first clinical component. This is what we expect because the example data is from the Y chromosome.

The figure created looks like:

From this figure, we can see that the genders separate nicely, corroborating what we see from the ANOVA p-value above. However, we can also see that there is a female clustering with the males. We can turn on labelling by adding -l Barcode to the command above to see the Barcode of the outlier and check the data for that individual.