Using PRISM¶
Quickstart¶
Install PRISM with pip.
$ pip install subclone-prism
Run analysis.
------------------------------
Extract epiloci from BAM file.
------------------------------
$ prism extract -i sample.bam -o sample.met
-------------------------------------------------------------------
Preprocess epiloci to get finer estimates of epigenetic subclones
and to rescue more fingerprint epiloci from noisy methylation data.
-------------------------------------------------------------------
$ prism preprocess -i sample.met -o sample.corrected.met
----------------------------------------------
Infer the subclonal composition of the sample.
----------------------------------------------
If you want 1-sample deconvolution, run:
$ prism deconvolute -i sample.corrected.met -o sample.prism.result
or if you want 2-sample deconvolution, run:
$ prism deconvolute -i sample1.corrected.met sample2.corrected.met -o sample.prism.result
--------------------------------------------
Scatterplot for visualization of the result.
--------------------------------------------
$ prism scatter -i sample.prism.result -o sample.png
--------------------------------------------------------------------
Annotation of fingerprint epiloci for functional characterization of
discovered epigenetic subclones.
--------------------------------------------------------------------
$ prism annotate -i sample.prism.result -o sample.prism.annotated.result \
--beds annotation_a.bed annotation_b.bed \
--annotation-names ANNOTATION-A ANNOTATION-B
The PRISM analysis is done in three steps: extract - preprocess - deconvolute.
Extract¶
Note
This section is under construction.
prism extract command extracts viable epiloci from a BAM file. In other words, it extracts genomic regions harboring a sufficient number of mapped reads (>= d) with a sufficient number of CpGs (>= c). A resulting file with those epiloci information is generated, whose file extension will be .met afterwards. To extract epiloci with default settings (d = 20, c = 4), simply run the command below:
$ prism extract -i sample.bam -o sample.met
If you want to specify your own cutoffs for the required sequencing depth and the number of CpGs, say, d = 15 and c = 3, run as follows:
$ prism extract -i sample.bam -o sample.met -d 15 -c 3
Preprocess¶
Note
This section is under construction.
prism preprocess command corrects for the errors in methylation patterns in order to amplify the number of fingerprint epiloci and calibrate for the subclone size estimates.
$ prism preprocess -i sample.met -o sample.corrected.met
You may benefit from multithreading with -t/--threads option.
$ prism preprocess -i sample.met -o sample.corrected.met -t 30
This step is resource intensive, so by default PRISM tries to pre-filter the epilocus that is not likely to be a fingerprint epilocus. This pre-filtering of course can be turned off by -f/--no-prefilter option and this indeed gives a better estimates of subclones. However, please be warned, depending on your data size, this step will last long. To help you deciding whether or not to apply prefiltering, with 30 threads (-t 30) ~200M met file took about 5 hours to be preprocessed without prefiltering.
$ prism preprocess -i sample.met -o sample.corrected.met --no-prefilter -t 30
For a more detailed description about all options, run prism preprocess -h.
Deconvolute¶
Note
This section is under construction.
prism deconvolute command infers the subclonal composition of the sample. Simply give methylation pattern-corrected epiloci file.
$ prism deconvolute -i sample.corrected.met -o sample.prism.result
Another feature of PRISM is that is can utilize two or more samples from a single tumor at the same time, and jointly infer subclonal composition. To provoke joint-analysis, specify a list of corrected.met files.
$ prism deconvolute -i sample1.corrected.met sample2.corrected.met -o sample.prism.result
For a more detailed description about all options, run prism deconvolute -h.
Scatter¶
Note
This section is under construction.
prism scatter command generates a scatterplot of the PRISM analysis result. You need a result of prism deconvolute. The dimension of analysis (i.e., the number of samples you gave to prism deconvolute command) should not be more than three to visualize it. Note that the file extension of output file should be a general one for image files such as png, jpg, or pdf.
$ prism scatter -i sample.prism.result -o sample.png
Annotate¶
Note
This section is under construction.