PMC

Distribution of mutations on the human tumour repressor TP53 with respect to their respective loss of information scores. The upper panel shows the distribution of D for 95 somatic, 49 cell-line and 15 germline NM in the TP53 tumour suppressor. The lower panel show the distribution of D for 1256 somatic, 127 cell-line and 36 germline FS indels reported for TP53 protein. These mutations are associated with a wide range of cancers.

Computation of the information loss score for variations in the protein-coding DNA. (a) Deletion of DNA base "A" at site 143 of the human protein PTEN causes a reading frame-shift (depicted by an arrow to the left on DNA sequence) that results in the mistranslation of all residues after the mutation. An alignment of homologous proteins is used to measure the biological information content of the reference sequence (PTEN in the figure) and to measure the loss of information due to this mistranslation. (b) A NM at the DNA site 910 on the human protein NF1 causes a premature stop codon at the residue 304 (denoted by *). The mutant protein is missing residues after the premature stop codon.

Computation of the LLR score for indels in the non-coding DNA. Once the indel is detected in population (here 39 S. cerevisiae yeast strains depicted in the top panel), an alignment of width W = 21 of orthologous sequences (referred to as the DNA site) from 5 closely related yeast species (including the reference S. cer, S. paradoxus, S. mikatae, S. kudriavzevii, and S. bayanus) is made centred around the position of the indel (bottom panel). The conservation of the DNA site is measured through the LLR score by measuring the relative rate of evolution of the site compared to the synonymous rate of evolution of the protein's coding region. (a) shows an insertion that falls into a highly conserved DNA site whereas (b) shows a deletion that falls into a region that is not conserved.

Spectrums of mutation allele frequencies. (a) Distribution of yeast frame-shifting indels (blue) skewed to lower DAFs compared to genes with in-frame indels (red) suggesting a relatively greater selective pressure on these genes. (b) Distribution of yeast NMs (blue) skewed towards the lower DAFs compared to the genes with synonymous SNPs (red) suggesting a relatively greater selective pressure. (c) Distributions of yeast FS indels that cause (blue) larger information loss (D > 0.3) and the those that cause (red) lower loss (D < 0.3). (d) In the case of genes with NMs, no mutations with higher allele frequency were observed and therefore the results are inconclusive. (e) Distribution of yeast non-coding DNA sites harbouring indels versus the indels' derived allele frequency (DAF) for two classes of indels: (blue) indels that fall in highly conserved sites (LLR > 16.6) and (red) indels that fall in less conserved sites (LLR < 16.6). The threshold THR = 16.6 is obtained using the Poisson random fields method (see Methods). (f) Distribution of human NMs versus their minor allele frequency in the human population for two classes of mutations causing (blue) greater information loss (D > 0.3) and (red) less information loss (D < 0.3), respectively.

Randomization experiments. (a) Distribution of the information loss caused by the FS indels in the yeast population (blue) compared to the density of a set of randomly distributed FS indels throughout the same set of yeast genes (red). (b) Distribution of the information loss caused by the NMs in the yeast population (blue) compared to a set of randomly distributed NMs throughout the same set of yeast genes (red). (c-d) Distributions of FS indels and NMs in the human population (dbSNP []), respectively, with respect to the information loss they cause. For each type of variation, this distribution is different when compared to the variations with records of disease association (green) and variations that do not have such records (blue). A set of randomly generated FS indels (red) shows a similar distribution to those that are associated with diseases. (e) The distribution of DNA sites that harbour indels with respect to their LLR score is compared to the distribution of DNA sites randomly chosen from 1000 bp upstream of all genes in the yeast dataset. In panels a and b, the "randomized" histogram bars represent the mean of 100 random samplings of the data, and the error bars represent the standard deviation observed over the 100 samplings, while in panels c-e the "randomized" histogram bars represent the mean of 50 random samplings of the data, and the error bars represent the standard deviation observed over the 50 samplings.