a. List of drugs relevant to Figures 1 and 2, and their possible metabolizing variants
b. Possible interactions between drugs and relevant variants

Figure 1: Susceptibility to Cancer
a. Schematic/Description of p53 variant
b. Other cancer-related variants in Myriad/Other screens
c. Cancer related variants in WGS based on GWAS (and comparison to NA12878 and other family members)
d. Cancer related variants based on TCGA (Alex's data)

Quick, simple brainstorm assuming the metabolomics data is q.c.'d and we are only looking at one run at a time.

** Patterns between metabolites

1. Identification of dense clusters of correlation
2. Comparisons of correlation among subpathways and across subpathways

** Patterns within individuals (time-course)

1. Univariate and multivariate change point analysis
2. Outlier detection
3. Fits of increasing and decreasing models
4. Associations with other datasets and eating habits
5. Creation of baseline levels

** Patterns across individuals

1. PCA and similarity scores across family members
2. Quantification of inter vs intra individual variance for metabolites/pathways/superpathways
3. Baseline levels modulated by observed changes in family

** Patterns across populations

1. PCA and similar scores across family members vs general population
2. Finer resolution of inter vs intra individual variance
3. Detection of outliers for tanner family based on population baselines

To look at differences in trends that might indicative of a change. Find changepoint that show a difference in mean or variance after a time point.

Use simple z-scores to find outliers across the time points.

Issue #4 is more about univariate cases with the changepoint package in R. However, just manually browsing through the graphs shows that there are trends. Packages such as http://arxiv.org/pdf/1309.3295.pdf allow for multivariate observations. Not sure if it will work on our data, because it's unclear to define if two metabolites levels should be correlated down, but worth exploring at the level of the SUB.PATHWAYS as defined by Metabolome.

Quantify extent of run variation:

• Variation in the number of metabolites detected
• Variation in the ordering or z-score of the metabolites within a sample
• Variation in the ordering of metabolites across samples

a. Overlap between other tools
b. Tool-specific likely pathogenic variants
c. Compound heterozygosity

Explore how change points differ depending on the run. Although Run 5 was a bit odd; would be good to compare since it had 6 of the 8 time points.

Quick brainstorm of RNA-Seq research plans.

We have 5 samples for the individual 002, 3 samples of individual 003, and then one for individuals 004 and 005.

Variation Across Individuals

• Comparisons across individuals to confirm that these samples are different people. Standard PCA plot.

Variation Within Individuals

• Overall trends (similar to PCA); ideally more differences across individuals than within individuals
• Changes over time
  • Linear Trends
  • Non-linear trends (change point analysis; spikes (outliers))

Comparisons with metabolomic data

• Difficulty level: over 9000
  • Studies by metablome showed that little to no correlation between blood expression and plasma metabolites, since plasma metabolites come from entire body.
• Metabolite - Gene relationships mined from http://consensuspathdb.org/
• Correlation across time
• Lag in response; although this is a month difference between timepoints, maybe metabolite level precedes increase in expression level. Although this will again be muddled.

Feel free to add or modify these ideas as you see fit.