SigTime output analysis result

mutational process

This represents the result of Sigprofiler.

Identifying the number, N, of mutational processes operative in
a set of cancer genomes is required prior to deciphering their
signatures. For every N, we evaluate the similarity between the
extracted processes (i.e., process reproducibility) from
stochastically initialized iterations.
For every N, we approach assesses the average Frobenius
reconstruction error of the averaged deciphered signatures P
and their strengths E. Low reconstruction error is indicative of
an accurate description of the original cancer genome catalogs.
We select the value of N for which the extracted
processes are reproducible and the reconstruction error is low.
In that set, the reconstruction error is the lowest when N is 3,
so the value of N was selected as 3.

Sigprofiler output(N=3)

Figure 1. Signature 1 of 3
Three mutational signatures deciphered from the base substitutions identified in samples.
Figure 2. Signature 2 of 3
Three mutational signatures deciphered from the base substitutions identified in samples.
Figure 3. Signature 3 of 3
Three mutational signatures deciphered from the base substitutions identified in samples.

Each Signature / COSMIC cosine similarity

image-ex
image-ex




timeseries analysis

As mentioned above, we used data deciphered into three signatures. Then, contributions to the signature for each mutation type were used to find out which signature was caused by more operative for each mutation.
For each signature, the sum of the signatures of the gene can be calculated using the deciphering. We used the analysis results calculated from DESeq, which showed logFC values measured at each time points. Using this, we can represent patterns of gene expression at time points.
As a result, we analyzed the timeseries by calculating the correlation between this pattern and each signature in samples.

timeseries analysis of SAMPLE1  =>   file

From the timeseries analysis of the samples, only the values satisfying 'correlation <= -0.9, p-value <= 0.05' were selected.

The results of node classification of the multi-labels of the analysis results into a single label are shown in the picture below.