cancer database on pyrimidine pathway

1 Final Conclusion

2 Background

2.1 denovo pyrimidine pathways

“…This supposition is corroborated by analyzing the Cosmic database, which document that genes of the pyrimidine de novo pathway are rarely mutated in cancer patients. More specifically, the rate of mutation of these genes in tumours is similar to that of non-cancerous tissues (Figure 6E.”

• Comparing mutation burdern in Cancer(somatic mut) and Normal (germline)
  • CAD, UMPS, DHODH, CPS1

2.2 Plans

• MR_recurrent/MR_nonrecurrent

• nonsyn/syn (kaks ratio)

• raw MR/ expression level.

2.3 Preparation

3 Figures

3.1 Overview

Table 3.1 contains basic information of the 4 genes in 1000Genome and TCGA cancer genome.

Table 3.1: Pyrimidine 4 gene mutation

Hugo_Symbol	Protein_length	CDS_length	Nmut_1000G	Nmut_cancer	mutrate_1000G	mutrate_cancer	Mr_1000G_scaled	MR_cancer_scaled
CAD	2225	6678	114	311	6.80e-06	3.9e-06	-0.4650436	-0.1719135
CPS1	1506	4521	70	402	6.20e-06	7.4e-06	-0.6440362	0.7353631
DHODH	395	1188	33	55	1.11e-05	3.9e-06	0.7420222	-0.1777953
UMPS	480	1443	23	65	6.40e-06	3.8e-06	-0.5926594	-0.2046028

Mutation rate \(MR_{gene}\) is normalized by sample size \(S\) and coding DNA sequence length\(L\). \(MR_{gene}\) is defined as follows:

\[\begin{equation} MR_{gene}=\frac{n_{gene}}{SL} \end{equation}\]

3.2 All cancer genes

• (not good), many cancer genes are not affected by somatic mutations. Figure 3.1

Figure 3.1: cancer genes as in CGC and uniprot(829 in total, )

3.3 Cancer genes mainly affected by SNVs and indels

• Only considering cancer genes with SNVs and small Indels, translocation, large amplifications/deletions mutation types were removed,resulting in 268 genes.

Figure 3.2: cancer genes as somatic, missense,frameshift, nonsense or splice site, excluding translocations, large amplifications/deletions

Figure 3.3: facets of previous figure

3.4 Cancer genes with recurrent mutations

We could further compare pyrimidine genes with cancer genes, which have at least 3 recurrent mutations(recurrence >=5)

Table 3.2: Gene type distribution

Var1	Freq
	6
Non-cancer	187
oncogene	17
oncogene/fusion	3
oncogene/TSG	6
onc/TSG/fusion	1
TSG	22
TSG/fusion	1

Table 3.3: Top 20 genes with number of recurrent mutations and gene type

Hugo_Symbol	N_recurr	Genetype
TP53	197	oncogene/TSG
PIK3CA	50	oncogene
APC	36	TSG
PTEN	35	TSG
TTN	24	Non-cancer
BAGE2	19	Non-cancer
CTNNB1	19	oncogene
CDKN2A	18	TSG
MUC4	16	Non-cancer
EGFR	14	oncogene
ARID1A	12	TSG
FBXW7	12	TSG
RB1	12	TSG
SMAD4	12	TSG
KRAS	10	oncogene
NFE2L2	10	oncogene/TSG
VHL	10	TSG
BRAF	9	oncogene/fusion
DNAH5	9	Non-cancer
ERBB2	8	oncogene/fusion

Figure 3.4: Comparing with recurrently mutated cancer genes, highlighting top5 in the above table and pyrimidien genes

4 Figures (updates)

4.1 Cancer genes with recurrent mutations (breast cancer specific)

• if we only include breast cancer mutations:
• DHODH is not shown because it doesn’t have mutations

Figure 4.1: With only breast cancer mutations

4.2 MR_recurrent/MR_nonrecurrent in TCGA and 1000G

4.3 MutsigCV q-value

Q value indicates mutation burden

4.3.1 Using breast cancer mutations only BRCA

4.3.2 Use 20 cancer types [pan20]

4.3.3 Barplot

• distribution

Table 4.1: pan cancer-20,distribution of qvalue < 0.1

	qvalue > 0.1	qvalue < 0.1
Non-cancer	18580	41
oncogene	58	7
TSG	70	20

4.3.4 Boxplot

4.4 dn/ds

• ratio of non-synonymous mutation rate/ synonymous mutation rates

4.5 raw mutation rate - expression level

4.5.1 BRCA

gene expression level (RNAseq V2 RSEM (log2))

## `geom_smooth()` using method = 'gam'

4.5.2 Raw MR- expression level

## `geom_smooth()` using method = 'gam'

4.5.3 Raw MR- expression level - highlighting recurrently mutated genes

## Joining, by = c("Hugo_Symbol", "Genetype")

## `geom_smooth()` using method = 'gam'

4.5.4 Highlight significantly mutated genes:

• 22 genes found to be significant in three or more tumor types:

  • TP53, PIK3CA, PTEN, RB1, KRAS, NRAS, BRAF, CDKN2A, FBXW7, ARID1A and MLL2, as well as STAG2

  • ATM, CASP8, CTCF, ERBB3, HLA-A, HRAS, IDH1, NF1, NFE2L2 and PIK3R1

## `geom_smooth()` using method = 'gam'

5 Supplements

5.1 Distribution of mutation rate

QQ plots of gene mutation rates(\(log_{10}(MR_{gene})\)) against normal distributaion in figure 5.1 and figure 5.2 appear to be linear(excluding the lower tails), which indicate that gene mutation rates follow a log normal distribution. So the mutation rates were transformed logarithmically, then they were scaled using z-score.

Figure 5.1: Q-Q plot of gene mutation rates in 1000G

Figure 5.2: Q-Q plot of gene mutation rates in TCGA cancer genome

5.2 Plain mutation rate

Figure 5.4 and 5.5 are reproduces of figure 5.3 (which used COSMIC mutation database)using TCGA mutation database.

Figure 5.3 doesn’t make much sense because I included all genes related to pyrimidine metabolism .

Figure 5.3: COSMIC(tumor) MR vs 1000G(normal) MR

Figure 5.4: TCGA(tumor) MR vs 1000G(normal) MR

Figure 5.5: a hexagonal heatmap representation of fig3

5.3 Cancer gene types

• Boxplots in 3.3 indicate that cancer genes with translocation roles in cancer have lower mutation rates than genes with SNVs and indels :

Figure 5.6: boxplot of MR translocation cancer genes and others

5.4 Gene non-synonymous/synonymous sites ratio

5.5 Gene expression distribution

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.