Behavioral Genomics Lab - R21 Gene-Expr proj / update 10.1.09


1. Subgroup Stratification


+ Stratification by demographic and clinical factor

[Blue DX=Bipolar; Red DX=Schizophrenic; Black DX=Control]
- note relative over-rep Females in Control subgr
- note relative over-rep "Substance use" among BP+SZ



2. Rank-ordered Genome-wide Expr Pattern Matrix "At a glance"


+ Expression levels for 21980 transcripts as they appear after rank-ordering (no other quantization aka binning) over the 30 subjects
ie, each cell of this 21980x30 matrix has values in the range [1,30] with lower ranks corresponding to under-expressed xscripts
[bluish = relative underexpression; reddish=relative overexpression, independently for each transcript;
vertical lines separate patient groups by DX]
- note simultaneous (relative) over- and/ -expr of "most" xcripts for pt #8(BP), #11(SZ), #28(CO)...Biological variability?
rank.ordered.expr.1.png [1-4999]
rank.ordered.expr.2.png [5000-9999]
rank.ordered.expr.3.png [10000-14999]
rank.ordered.expr.4.png [15000-19999]
rank.ordered.expr.5.png [20000-21980]



3. Rank-ordered Expr Histograms

+ Each curve is a 30-bin histogram corresponding to an individual in the study. Histograms are constructed by counting the # of transcripts for an individual across ranks 1-to-30.
example histogram (later curves are more graphically efficient allowing superposition of many individual's histograms)


[by DX]
There appear to be 2 stylized types of individuals:
- "extreme"-type individuals have a high count of transcripts that are over-expressed (relative to n=30 sample) and simultaneously an approximately similar count of (distinct) transcripts that are under-expressed (relative to n=30 sample)
- "centered"-type individuals have a relatively even count of transcripts at intermediate ranks (around the median) that tapers off at the min and max ranks (relative to n=30 sample)
- It's unclear why this dichotomy appears at the genome-wide analysis level, seemingly independently of DX or Gender subgroups. Either it is of real biological origin or Partek's normalization algorithms are distorting the underlying gene expression "scan-level" datasets


[by Gender]
- Unbalanced M/F subgroups (21/9) and the existence of the 2 stylized expression patterns potentially explains why males seem to, as a group, display "wider" expr patterns


[By HX Psychosis]


[By Smoke]


[By Stylized-expr patterns]
This subgroup partition cleanly discrimintes the two stylized expr patterns, but what do these individuals have in common?





4. Rank-ordered select-xscrip Histograms

scores are based on median-differences normalized by interquartile-range