A tool for timing complex copy number gains in cancer. Provides gain timing estimates for segments with a total copy number of up to 9. Only copy number segments with 10 or more SNVs will be timed.
Each gain timing is measured in mutation time, a scale that ranges from 0 to 1. A timing of 0 indicidates that the gain occured close to conception and 1 that the gain occurred very close to the emergence of the tumour's most recent common ancestor.
GRITIC can be installed using pip
pip install gritic
The easiest way to run GRITIC on a single sample is to use the rungritic_cmd.py script.
python rungritic_cmd.py -ARGS
This script has five required arguments.
--mutation_tableA path to the SNV table for the sample.--copy_number_tableA path to the copy number table for the sample.--purityThe estimated cellular purity for the sample.--sample_idSample ID, for output purposes only.--outputThe directory for the output, if it doesn't exist it will be created.
There are also a number of optional arguments.
--subclone_tableA path to the subclone table for the sample. If not provided it is assumed every SNV is clonal.--wgd_statusGRITIC will identify the WGD status of the sample. This can be overidden through this argument.--non_parsimony_penaltyApply a penalty on non-parsimonious routes in WGD tumours. Defalt is False. See preprint for more details.--plot_treesPlot the route trees for each segment. Default is True.
An command to get GRITIC to run using the example data is:
python rungritic_cmd.py --mutation_table examples/snv_table_example.tsv --subclone_table examples/subclone_table_example.tsv --copy_number_table examples/cn_table_example.tsv --purity 0.5 --wgd_status T --output examples/output --sample_id TEST_ID
Alternatively, the similar script rungritic.py provides a very simple framework to run gritic as part of a larger script.
The three input tables that GRITIC requires should be tab separated. Examples using simulated data are available in the example directory. We currently filter any non-autosomal chromosomes.
All SNVs for the sample. Requires the column names Chromosome, Position, Tumor_Ref_Count & Tumor_Alt_Count.
The rounded allele-specific copy number profile for the sample. Requires the column names Chromosome, Segment_Start, Segment_End, Major_CN & Minor_CN. Major_CN must be larger than or equal to Minor_CN.
GRITIC will merge adjacent segments with the same allele specific copy number.
The identified subclonal peaks and their relative sizes for the sample. All peaks with a cancer cell fraction of less than 0.9 should be included. Any peaks higher than this are automatically filtered. GRITIC assumes a peak with a cancer cell fraction of 1.
This table requires the column names Subclone_CCF (the cancer cell fraction of the subclone) & Subclone_Fraction (the fraction of total SNVs present in the subclone). An Cluster column is also required as an index for the subclones.
Only subclones with more than 10% of the subclonal SNVs are included. If there are more than two subclones, GRITIC will reformat the sample to have two subclone. The subclone with the largest CCF is unmodified and the remaining clones grouped together. This speeds convergence while maintaining an accurate estimation of clonal vs subclonal SNV composition.
This table is optional, if it is not included GRITIC will assume every SNV is clonal. This will likely bias the gains to be measured earlier.
GRITIC produces a number of outputs to describe the timing of copy number gains in a given tumour:
Gives a summary of the gain timing infotmation for each gained segment. For each gained segment, the sequential gains are labelled with Gain_Index. The median and the 95% percentiles for the gain timing for each sequential gain are reported.
In WGD tumours, the number of gains that arise independently of the WGD will vary depending on the route. Only gains that arise independently of the WGD in at least 80% of posterior samples are reported, this is recorded in the Proportion column. The timing of the WGD and the probability that each gain arose before the WGD is also recorded.
This table gives 250 direct samples from the timing posterior for each gained segment. Both the gain and the WGD timing are given in the case of a WGD tumour.
Each individual sample from the posterior for a given segment is labelled with Posterior_Sample_Index. The route for each posterior sample is given by the route column.
Again, as the number of gains that arise independently of the WGD will vary depending on the route, the number of gains per Posterior_Sample_Index can vary in WGD tumours.
This table gives more details about each possible route for each gained segment. As well as giving the timing for each gains in the route, it also records the route probability that is used when sampling the posterior. The total number of events implied for each route is also given, as well as the density of the timing samples.
Please see the supplementary materials of the accompanying preprint for more details on the route densities.
A table with a single row that gives WGD calling information for the sample.
If the most common major copy number by total base pair length (Major_CN_Mode) is 1 then the sample is non-WGD. If the Major_CN_Mode is > 2 then the sample has likely undergone multiple WGDs, but this history is not currently supported by GRITIC.
If the Major_CN_Mode is 2 then a secondary test is conducted to test the WGD status. All major copy number two segments are timed and the point in mutation time that overlaps with the greatest proportion of segments by width is recorded. If this is greater than 60% then the sample is identified as WGD. The best overlap timing and the proportion are recorded in the columns Best_Overlap_Timing and Overlap_Proportion respectively. The timing of the WGD is then jointly estimated from the overlapping segments and recorded in the Timing,CI_High and CI_Low columns.
This table is produced for WGD tumours. It gives the timing of the major copy number two segments which are assumed to primarily arise through a WGD. The Best_Overlap_Timing column records the point in mutation time that overlaps with the 89% confidence intervals of the timing of the highest proportion of basepair length of major copy number two segments. The proportion is recorded in Proportion and whether the given segment's timing confidence intervals intersect with the timing by Intersecting.
The mutation table processed by GRITIC to give additional SNV multiplicity information. SNV multiplicity probabilities are given by the Prob_Mult_ columns. The expected number of mutations with the same coverage as the given SNV that would be expected to have less than three reads supporting the alternate allele is given by the Three_Reads_Correction_Mult columns. This is approximately the number of SNVs that would be missed by the variant caller at this coverage level.
Binary tree plots for the gain timings for each route for a given segment. Each plot has two binary trees corresponding to each parental allele. Blue nodes represent independent gains, yellow nodes WGD gains and red nodes the final alleles that were present at time of sampling.
Python dictionary objects containing the stored gain timing and multiplicity posterior samples for each route for each gained segment. Not necessary for most use cases. Can be read using the pickle and bz2 modules in python. For example:
import pickle
import bz2
with bz2.BZ2File('SAMPLE_ID_timing_dicts/SEGMENT_ID_timing_dict.bz2', 'rb') as f:
timing_dict = pickle.load(f)
The keys of each dict correspond to the routes for the sample. Within each route there are Timing and Mult keys. The Timing entry gives the timing samples of the independent gains indexed by the corresponding node in the tree. A WGD timing entry is also given if applicable.
The Mult entry gives the multiplicity proportions corresponding to each timing sample. It is a N_SamplesxN_Multiplicities numpy array. Across the columns, the multiplicities are orderred from 1 to the major copy number of the segment, followed by the subclonal multiplicity probabilities.