Chai-squared test in R

In this study (1), I was involved with a part of research that analyzes the function of Conserved Noncoding Elements (CNE) upstream of Amphioxus Msx gene in amphioxus development.

The fragment of Msx-CNE was amplified by PCR and inserted into the Lac-Z expression vector, containing the minimal promoter (including TATA box, CCAT box and GC box element) of amphioxus FoxD. I don’t mention the episode of plasmid construction here. About this, see other article.

Although negative control (no enhancer: LacZ) was not supposed to express LacZ, 0-3 animals out of several dozens of animals had small lacZ expression in the body (Above pictures K, N). Needless to say, the injection of MSX-CNE caused strong lacZ expression (Above pictures I、J、L、M). Whether injections went well was confirmed by checking the fluorescence of dextran. Therefore, I guess that the negative control plasmid didn't work in animals have no lacZ expression. However, after discussion, we decided to counted the number of animals that have lacZ expression.

For statistic analyzes, we had a discussion, like....

Boss: What do you think how many samples are enough for the statistic analyze?

Me: Well, I'd say, more than 30 animals...50 animals are enough.

Boss: K, collect more than 50 animals.

Me: I see. (I should have said just "more than 30". I thought that in my mind.)

After all, I collected 59 animals that is injected negative control plasmid and having lacZ expression. What a nice enough number of animals!! 

Finally I could start a statistic analyze!

To examin whether there is difference between negative control (but lacZ expression exists) and Msx-CNE in expression of lacZ, I counted the each number of lacZ expression in ectoderm, somite, notochord, nerve cord (neural tube), endoderm and abnormal cells in gut (They are transient cells that will be disappeared in later stage. Probably they are derived from accidentally detached endodermal cell.). I aggregated the number and wrote in notebook, and then made a Excel file (don't forget to save the file as a .csv file). 

1. Save the .csv file on the desctop.

2. Specify where you run R. (For Mac)

> setwd("/Users/Username/Desktop") 

> getwd()　　3. Confirm where you run R. Nothing is needed in the ( )

［1］"/Users/Username/Desktop"

You can see where you run R.

4. Read csv file.

> MSX<-read.csv("MSXsheet.csv")

>MSX　5.Open the file. 

1 means there is/are expression(s) and 0 means there is no expression. The reason why both 1 and 0 are needed is the information total number of animals (59 animals) is needed for the analyze. 

6. Making table (if you make the table of plasmid x ectoderm).

> table(MSX$PLASMID,MSX$ECTODERM)

7. Calculate by chi-squared text. The command is..

 > chisq.test(MSX$PLASMID,MSX$ECTODERM, correct=FALSE)

The P-value is larger than 0.05. It's not significant. 

For other structure..

All are not significant.

Now, how about "somite"???

The P-value is 0.0000002928!!

It would be suggested that both negative control and MSX-CNE have an ability to express lacZ in any tissue potentially but MSX-CNE could change the activity to express in the somite dominantly.

As far as I know, Msx gene is expressed in the neural plate border region in amphioxus. We couldn't observed the expression in this study. Probably another enhancer region would have a role in the expression.

References

1. Yue, J-X.; Kozmikova, I.; Ono,H.; Nossa, C. W.; Kozmik, Z.; Putnam,N. H.; Yu, J-K.; Holland L. Z. (2016). "Conserved Noncoding Elements in the Most Distant Genera of Cephalochordates: The Goldilocks Principle". Genome Biol. Evol. 8(8):2387-2405.

2. Yu, J-K.; Holland, N. D.; Holland, L. Z. (2004). "Tissue-specific expression of FoxD reporter constructs in amphioxus embryos". Developmental Biology 274:452–461.

4. 栗原伸一. (2011). "入門統計学ー検定から多変量解析・実験計画法までー". オーム社.

3. 村井潤一郎.(2013). "はじめてのRーごく初歩の操作から統計解析の導入までー". 北大路書房.

➢