Mutational signatures in cancer with DNA-Seq

A recent collaboration with a clinician here at UI hospital and clinics introduced me to the idea of mutational signatures in cancer.  Characterizing mutational signatures is made possible by the falling cost and increasing accuracy of whole-genome sequencing methods.  Tumors are sequenced across the entire genome and the catalog of somatic mutations (i.e, SNPs) is used to compute the mutational signatures of a tumor’s genome.

The idea is that the collection of somatic mutations found in a tumor  are the result of a variety of defective DNA-repair or DNA-replication machinery combined with the action of known or unknown mutagens and environmental exposures.  The processes operate over time and leave a “footprint” in the tumor DNA that can be examined.  These sum of all of the mutational processes operating within a tumor cell is a distinct mutational “signature” that differs by tumor types.

For example, in lung cancer, the bulk of somatic mutations are C>A transversions resulting from chronic exposure to tobacco smoke.  In melanoma, the predominant mutation type is C>T and CC>TT at dipyrimidines, a mutation type associated with UV-light exposure.  And in colorectal cancer, defective DNA mismatch repair contributes the majority of the mutations.

Mutational signatures of a cancer by the operation of several mutational processes over time.
Mutational signature of a cancer by the operation of several mutational processes over time. From Used under CC License BY 3.0.

A recent paper in Nature has formalized this notion of mutational signatures in tumors and provided a mathematical framework (written in MatLab) for assessing how many and which signatures are operational within an uncharacterized tumor type (generally there between 2 and 6 processes).

In the paper, the authors analyzed almost 5 million somatic cancer SNPs and identified 21 unique signatures of mutational processes through a mathematical process of deconvolution, followed by experimental validation.  A curated catalog of the most current signatures based on available sequence data can be found at the COSMIC database.

In part 2 of this post, I’ll go into more detail on the mutational signatures and link to some python code I’ve written to help get flat-file lists of SNPs into the correct form for easy input into the MatLab framework.




Exploratory analysis of human splice-altering variants

Single splice-altering variants can alter mRNA structure and cause disease

The splicing of introns and joining of exons to form mRNA is dependent on complex cellular machinery and conserved sequences within introns to be performed correctly.  Single-nucleotide variants in splicing consensus regions, or “scSNVs” (defined as −3 to +8 at the 5’ splice site and −12 to +2 at the 3’ splice site)  have the potential to alter the normal pattern of mRNA splicing in deleterious ways.  Even those variants that are exonic and synonymous (i.e., they do not alter the amino acid incorporated into a polypeptide) can potentially affect splicing.  Altered splicing can have important downstream effects in human disease such as cancer.

Using machine-learning to predict splice-altering variants

In the paper “In silico prediction of splice-altering single nucleotide variants in the human genome,”  the researchers took on the problem of predicting which single-nucleotide variants (SNVs) have the potential to be splice-altering by computing “ensemble scores” for potential variants, combining the results of several popular splicing prediction software tools into one probability score.

They did this by using “random forest” (rf) and “adaptive-boosting” (adaboost) classifiers from machine-learning methods to give improved ensemble predictions that are demonstrated to do better than predictions from an individual tool, leading to improvements in the sensitivity and specificity of the predictions.

As part of their supplementary material, the authors pre-computed rf and adaboost scores for every SNV in a library of nearly ~16 million such sites collated from human RefSeq and Ensembl databases.   The scores are probabilities of a particular SNV being splice-altering (0 to 1).

Exploratory analysis of the database

I performed an exploratory data analysis of chromosome 1 (chr1) SNVs from the  database that was made available with the paper.

First, I just looked at where the SNVs on chrom 1 were located as classified by Ensembl region:

Fig 1. Number of variants in the scSNV database on chromosome 1  by Ensembl region. Not surprisingly, ‘intronic’, ‘exonic’, and ‘splicing’ are the most common regions for potential splice-altering SNPs.

As can be seen from Fig 1, most of the SNVs are located in introns, exons, and splicing consensus sites according to their Ensembl records.

Next, I created histograms for the chrom 1 SNVs by their Ensembl classification, looking at rf scores only (keep in mind that the scale on the  y-axis for the plots in Fig 2 and 3 differs dramatically between regions).  The x-axis is the probability of being splice-altering according to the pre-computed rf score.

Fig 2. Random-forest (rf) score by ensembl region for the ~15 M scSNVs in the database.

I noticed the fact that within ‘exonic’ regions on chrom 1, the rf scores take on a range of values from 0.0 to 1.0 in a broad distribution, while in other regions like ‘UTR3’, ‘UTR5’, ‘downstream’, etc… the distributions are narrowly skewed towards zero.  For the ‘intronic’ region, the majority of sites have low probability of being splice-altering, while at the ‘splicing’ consensus sites, the vast majority are predicted to be splice-altering variants.  This appears to make intuitive sense.

I performed the same analysis for the adaboost scores, as shown in Fig 3 (below).  You can see that the adaboost scores take on a more binary distribution than the rf scores, with any individual SNV likely to be classified as ~1 or 0 according to the adaptive boosting method.  Just like the rf scores, SNVs in ‘exonic’ regions are equally likely to be splice-altering as not, while those in ‘splicing’ regions are highly likely to be splice-altering.  An SNV in an ‘intronic’ regions is ~3X more likely to have no effect on splicing.

Fig 3. Ada score by Ensembl region for the scSNV database.


Finally, I looked at the relationship between the two scoring methods for the SNVs that fall within the Ensembl-characterized ‘splicing’ regions on chrom 1.  That scatter plot is shown below in Fig 4.

I suppose I was expecting a tight linear correlation between the two approaches, however the data show that the rf and adaboost methods differ substantially in their assessment of the collection of SNVs in these regions.

It is obvious from the plot below that there are many SNVs that the rf method considers to have low probability of being splice-altering that are found to have very high (>0.9) probability by the adaboost method.

Fig 4. Scatter plot of rf score vs. ada score for SNVs within 'splicing' regions on chrom 1.
Fig 4. Scatter plot of rf score vs. ada score for SNVs within ‘splicing’ regions on chrom 1.

This result would appear to suggest that if one is going to classify variants as “splice-altering” from this database, it would be best to consider both predictions or some combination of them rather than relying on either score alone if the goal is not to miss any potentially important sites.  Conversely, if the goal is to only consider sites with very high likelihood of being splice-altering, a threshold could be set such that both scores need to be above 0.8, for example. article on crispr/cas9 gene-editing technology

Wired is out with a great article covering the development and innovation of crispr/cas9 technology and its intertwined promise and peril.

Sample graf:

The stakes, however, have changed. Everyone at the Napa meeting had access to a gene-editing technique called Crispr-Cas9. The first term is an acronym for “clustered regularly interspaced short palindromic repeats,” a description of the genetic basis of the method; Cas9 is the name of a protein that makes it work. Technical details aside, Crispr-Cas9 makes it easy, cheap, and fast to move genes around—any genes, in any living thing, from bacteria to people. “These are monumental moments in the history of biomedical research,” Baltimore says. “They don’t happen every day.”

Using the three-year-old technique, researchers have already reversed mutations that cause blindness, stopped cancer cells from multiplying, and made cells impervious to the virus that causes AIDS. Agronomists have rendered wheat invulnerable to killer fungi like powdery mildew, hinting at engineered staple crops that can feed a population of 9 billion on an ever-warmer planet. Bioengineers have used Crispr to alter the DNA of yeast so that it consumes plant matter and excretes ethanol, promising an end to reliance on petrochemicals. Startups devoted to Crispr have launched. International pharmaceutical and agricultural companies have spun up Crispr R&D. Two of the most powerful universities in the US are engaged in a vicious war over the basic patent. Depending on what kind of person you are, Crispr makes you see a gleaming world of the future, a Nobel medallion, or dollar signs.

The technique is revolutionary, and like all revolutions, it’s perilous. Crispr goes well beyond anything the Asilomar conference discussed. It could at last allow genetics researchers to conjure everything anyone has ever worried they would—designer babies, invasive mutants, species-specific bioweapons, and a dozen other apocalyptic sci-fi tropes. It brings with it all-new rules for the practice of research in the life sciences. But no one knows what the rules are—or who will be the first to break them.