In a recent British Journal of Cancer article the authors have performed a preliminary analysis of genetic screening of those for higher risk for prostate and breast cancers. We herein look at the prostate cancer issue.
Simply stated the authors have assembled a database of genetic samples and for each have detailed the relative risk and the prevalence. Specifically:
1. They listed SNPs from the dbSNP (“Single Nucleotide Polymorphism database”). A SNP is a DNA sequence variation with a single nucleotide, ATGC, and may be in an exon or intron. Many of these variations occur.
2. The odds ratio, OR, is the odds of an event occurring in one group as compared to another. Thus we can say that if we have two groups, say group 1 which has the SNP alteration, and Group 0 which does not have the alteration, then the odds ratio is given by:
[p1/(1-p1)]/[p0/(1-p0)]
and if the odds ratio is greater than one then we have a greater chance of occurrence. Now consider two SNPs, and their respective individual and total odds ratio. Let p1 be SNP1 and p2 SNP2 and p0 be the lack of SNP1 and p00 the lack of SNP2. Then we have an odds ratio for both occurring, if independent, as:
[p1p2/(1-p1p2)]/[p0p00/(1-p0p00)]
This assumes independence and shows that the OR do not readily allow direct and simple calculation from each other separately. We of course can extend this principle to n SNPs. It is obvious
3. Using the SNPs as a measure of increased or decreased risk, one can set a risk threshold and test those above and ignore those below.
The result is given by the authors as:
Compared with screening men based on age alone (aged 55–79: 10-year absolute risk >2%), personalized screening of men age 45–79 at the same risk threshold would result in 16% fewer men being eligible for screening at a cost of 3% fewer screen-detectable cases, but with added benefit of detecting additional cases in younger men at high risk.
Similarly, compared with screening women based on age alone (aged 47–79: 10-year absolute risk >2.5%), personalized screening of women age 35–79 at the same risk threshold would result in 24% fewer women being eligible for screening at a cost of 14% fewer screen-detectable cases.
Similarly, compared with screening women based on age alone (aged 47–79: 10-year absolute risk >2.5%), personalized screening of women age 35–79 at the same risk threshold would result in 24% fewer women being eligible for screening at a cost of 14% fewer screen-detectable cases.
Personalized screening approach could improve the efficiency of screening programs. This has potential implications on informing public health policy on cancer screening
That is, by performing SNP analysis and ten establishing a threshold one can bifurcate the groups. One could also select groups in some graded multi-sector grouping as well.
The SNPs chose are shown in a modified form below. Many are on the same gene segment. There were a total of 31 SNPs as of the date of the paper where the odds ration exceeded 1.0.
The procedure here is an interesting first step in the genetic testing of potential cancer patients. The process however will most likely require significant refinements. The process however will most likely require significant refinements.
A simple approach to the above is to note that if we rank order the SNPs from highest OR to lowest, and then plot total OR versus N, total SNPs used as ranked, and then also the frequency of the same SNPs as ranked we get an interesting curve. I will leave that thought for the moment.
dbSNP No. | Locus/gene | Risk allele frequency | Odds Ratio per allele |
rs12621278 | 2q31/ITGA6 | 0.940 | 1.300 |
rs721048 | 2p15 | 0.190 | 1.150 |
rs1465618 | 2p21/THADA | 0.230 | 1.080 |
rs2660753 | 3p12 | 0.110 | 1.180 |
rs10934853 | 3q21.3 | 0.280 | 1.120 |
rs7679673 | 4q24 /TET2 | 0.550 | 1.090 |
rs17021918 | 4q22/PDLIM5 | 0.660 | 1.100 |
rs12500426 | 4q22/PDLIM6 | 0.460 | 1.080 |
rs9364554 | 6q25 | 0.290 | 1.170 |
rs6465657 | 7q21 | 0.460 | 1.120 |
rs10486567 | 7p15 /JAZF1 | 0.770 | 1.120 |
rs2928679 | 8p21 | 0.420 | 1.050 |
rs1512268 | NKX3.1 | 0.450 | 1.180 |
rs620861 | 8q24 | 0.610 | 1.280 |
rs10086908 | 8q24 | 0.700 | 1.250 |
rs445114 | 8q24 | 0.640 | 1.140 |
rs16902094 | 8q24 | 0.150 | 1.210 |
rs6983267 | 8q24 | 0.500 | 1.260 |
rs16901979 | 8q24 | 0.030 | 2.100 |
rs4962416 | 10q26 /CTBP2 | 0.270 | 1.170 |
rs10993994 | 10q11/MSMB | 0.240 | 1.250 |
rs7127900 | 11p15 | 0.200 | 1.220 |
rs7931342 | 11q13 | 0.510 | 1.160 |
rs4430796 | 17q12 /HNF1B | 0.490 | 1.240 |
rs11649743 | HNF1B | 0.800 | 1.280 |
rs1859962 | 17q24.3 | 0.460 | 1.240 |
rs2735839 | 19q13/KLK2,KLK3 | 0.850 | 1.200 |
rs8102476 | 19q13.2 | 0.540 | 1.120 |
rs5759167 | 22q13 | 0.530 | 1.160 |
rs5945619 | Xp11 | 0.280 | 1.120 |
A simple approach to the above is to note that if we rank order the SNPs from highest OR to lowest, and then plot total OR versus N, total SNPs used as ranked, and then also the frequency of the same SNPs as ranked we get an interesting curve. I will leave that thought for the moment.
Thus we can ask the questions as follows:
1. Which SNPs, say the set of some n of them, provides the best set to minimize mortality and minimize the number requiring testing?
2. Can there be some clustering of SNPs such that there are disjoint classes of individuals which get assigned to risk groups. Those in the highest receiving the most significant attention and those in the lowest receiving minimal?
3. Are the SNPs such that they are independent predictors or are there environmental or other exogenous factors which can effect SNPs alone?
4. What is the relationship between SNPs and the pathways known as part of PCa development?
5. Are there temporal changes in SNPs and is there some relationship between these temporal changes? Namely are there causal SNP changes?
6. What are the causes of the SNPs?
7. Knowing the SNPs and those with PCa, what can be determined regarding the dynamics of PCa development?
8. What is the relationship between SNPs and the prostate cancer stem cell? Does the CSC have different expressions?
Posts
