INVERSIÓN EXTRANJERA a. Marco Legal

GD 11 3215 2939 GD 12 3691 2946 GD 14 3971 3201 GD 16 3175 2431 GD 18 3578 3079 PN 2 3416 3188 500 1500 2500 3500 n (t ) DE − and n (t ) DE+

Time points t for the ovary development data set

Figure 6.6: Bar chart of the total numbers of significantly down (n(t)_DE−) and up

(n(t)

DE+) regulated genes with respect to the mixed reference measurement at postnatal

day PN 0.5 controlling a TS-ABH FDR of 0.05 for the embryonic ovary development data set.

The GEO series GSE5334 includes 19 raw data .cel-files of hybridized Affymetrix mouse 430 2.0 arrays. The study was conducted to learn more about the molecular

biological processes in the ovary development. The RNA from the pair of ovaries from three different female mice was hybridized on the microarrays at each of the six time points: gestational days GDs 11, 12, 14, 16, 18 and postnatal day PN 2. A single chip was hybridized as reference with a mix of the whole body RNA of four female and four male mice at approximately 12 hours after birth. Unfortunately, there is no original contribution, which reports results originating from this data set. The analyzed tissue in this experiment is narrowly defined, which allows to focus on the gene set activation profiles closely related to the ovary and its biological processes.

Figure 6.6 shows the absolute numbers of differentially expressed genes per time point and direction of regulation. Differential expression is determined with the

Dshrink

g statistic (see section 4.2) while controlling the TS-ABH FDR across all tests at a level of αgenes = 0.05. The number of significantly up and down regulated genes is clearly above 2000 across all time points as expected in a development experiment in contrast with a single static measurement (reference after birth). The numbers of differentially expressed genes are high in comparison with the other experiments analyzed in this contribution. The large number of differentially expressed genes must not inevitably lead to an extraordinary large number of a priori defined gene sets with a significant activation profile, although this is observed in the following. Table 6.6 reveals the most frequent activation profiles resulting from a 2S-GSA profile analysis combined with an AM smoothing with parameters λfill = 0.0462 and λwipe= 11.2813. The TS-ABH FDR is controlled at level 0.01 across all enrichment

Table 6.6: The 14 most frequent 2S-GSA activation profiles after smoothing with

AM (λfill = 0.0462, λwipe= 11.2813) resulting from the OD data set.

Profiles # Profiles # Profiles #

oooooo 5918 -ooooo 33 +ooooo 21

--- 478 ---ooo 30 ooo--- 21

++++++ 256 +++ooo 27 --oooo 18

++oooo 55 ooooo- 25 +++oo+ 16

tests. The maximum frequency among the 1300 non-zero activation profiles is achieved by the continuous down regulated profile (---) followed by the profile where all time points show a significant enrichment with significantly up expressed genes (++++++). Generally, profiles with a continuous segment are in the majority in contrast to the results from the AH profile analysis. Those continuous profiles are very likely not generated by chance and hence provide a more reliable insight in the gene set activity during the examined time series experiment than single position or irregularly alternating profiles do.

The large number of non-zero activation profiles makes it difficult to provide a deep insight in the analysis results. Table 6.7 lists 26 genes sets – ten to each extreme of the ranking according to the Dmed

s scores from section 4.4 and another six hand-picked from the STEM results. The top 100 according to the D|med|

s score are recorded in Table B.8 in the Appendix (the complete result table is available in the electronic version of the document). There are six GO gene sets whose term description reveals a direct link to the female reproductive system among the ten top ranked gene sets with up regulated positions in their profiles. Since also piRNA is closely related to the germ line cell (Lau, Seto, et al. 2006) this number can be enlarged to seven. The top ranked gene set with up regulation GO:0007066 includes only four genes, has the term description: “female meiosis sister chromatid cohesion” and its enrichment with significantly up regulated genes turns out to occur already at gestational day 14, whereas the other reproduction related gene sets become active after birth or in case of GO:0009566 (“fertilization”) at GD 18. The appearance of the well-fitting gene sets on a high position in the ranking emphasizes the adequacy of both the activation profile algorithm and the ranking method. This observation gives confidence that new findings are based on the underlying molecular genetic processes and not by chance, although there is no biological view in form of an original contribution based on this data set available for comparison purposes. The top ranked gene sets are relatively small (< 100) with some exceptions, which was a general finding in all data sets due to the score calculation (see section 6.2). The majority of significant sets is characterized with an activation profile with one or more positions enriched with down regulated genes and the ten top ranked profiles

of this class in Table 6.7 obtain in summary a higher rank due to the D|med| s score than the profiles with up regulated positions. Interestingly, among the top ten sets with a down regulated position there are only three GO sets, but three Reactome sets and four sets from the BioCarta gene set definition. The extraordinary high total number of continuously down and up regulated profiles across all examined time points can be explained with the type of reference. The mixed-tissue reference sample was taken after the embryonic development process finished and hence the molecular genetic differences must be larger than if the reference had been another embryonic tissue (e.g. brain or muscle). One of the 20 top ranked gene sets was also identified by the maSigFun procedure (GO:0034587 “piRNA metabolic process”), whereas the STEM algorithm reports rather larger gene sets as referred in section 6.2. Six of these gene sets, which are also identified by all GSA-type algorithms, are added to the top ranked profiles in Table 6.7. The GO term descriptions are mainly related to general developmental issues like proliferation, migration or motility. The largest gene set – GO:0048734 – listed with 2646 genes encompasses the quite general description “system development”.

The analysis of the ovary development experiment shows that the enrichment method is able to identify appropriate gene sets with a reasonable profile and the applied ranking supports the researcher by the identification of meaningful profiles even when the list of significant findings is large.

Table 6.7: Top 10 of Dmed

s -ranked up regulated and top 10 down regulated gene

set activation profiles resulting from a 2S-GSA profile algorithm combined with AM smoothing on the OD data. The six hand-picked additional sets at the bottom show interesting gene sets with lower ranking.

ID |med D

|

s

Rank

Profile Description |s| 2T-GSA 1S-GSA STEM maSigF

un

GO:0007066 10 oo+ooo female meiosis sister chro- matid cohesion

4 7 X 7 7

GO:0015671 60 o+oooo oxygen transport 6 7 X 7 7

GO:0048608 71 ooooo+ reproductive structure de- velopment

202 7 X 7 7

GO:0009566 74 oooo++ fertilization 82 7 X 7 7

GO:0022602 75 ooooo+ ovulation cycle process 74 7 X 7 7

GO:0042698 76 ooooo+ ovulation cycle 78 7 X 7 7

GO:0048511 79 ooooo+ rhythmic process 160 7 X 7 7

GO:2000194 83 ooooo+ regulation of female go- nad development

6 7 X 7 7

GO:0007130 104 oo++++ synaptonemal complex as- sembly

13 X X 7 7

GO:0034587 108 oo++++ piRNA metabolic process 9 7 X 7 X REACT:

COMMON

11 --- genes involved in common pathway

13 X X 7 7

BioCarta: INTRINSIC

9 --- intrinsic prothrombin ac- tivation pathway

16 X X 7 7

BioCarta: EXTRINSIC

8 --- extrinsic prothrombin ac- tivation pathway

13 7 X 7 7

. . . continued from previous page

ID Rank Profile Description |s| 2T-GSA 1S-GSA STEM maSigF

un

BioCarta: AMI

7 --- acute myocardial infarc- tion

15 X X 7 7

GO:0050790 6 o-oooo regulation of catalytic ac- tivity

1427 7 X 7 7

BioCarta: FIBRINOLYSIS

5 --- fibrinolysis pathway 10 7 X 7 7

GO:0042542 4 --oooo response to hydrogen per- oxide

52 7 X 7 7

REACT:GRB2 2 --oooo genes involved in GRB2 11 7 X 7 7 REACT:

P130CAS

3 --oooo genes involved in p130Cas linkage to MAPK signal- ing for integrins

11 7 X 7 7

GO:0070301 1 -ooooo cellular response to hydro- gen peroxide

30 7 X 7 7

GO:0003002 1216 o+oooo regionalization 285 X X X 7

GO:0008283 867 oo---- cell proliferation 1207 X X X 7

GO:0016477 952 oo---- cell migration 695 X X X 7

GO:0030154 907 oo---- cell differentiation 2202 X X X 7

GO:0048870 888 oo-oo- cell motility 750 X X X 7

GO:0048731 719 o--- system development 2646 X X X 7