4
$\begingroup$

I recently read the paper “A novel long non-coding RNA linc-ZNF469-3 promotes lung metastasis through miR-574-5p-ZEB1 axis in triple negative breast cancer”. In this I see Table1 showing correlation of Linc-ZNF469-3 with different features in TNBC and non-TNBC patients. p-values are calculated based on Pearson's Chi-sq test.

enter image description here

For non-TNBC patients: I calculated the p-value with the Chi-sq test just to have a look if everything is right.

Recurrence:

data <- data.frame(x= c(40,62), y=c(56,65))
data
chisq.test(data, correct = F)

    Pearson's Chi-squared test

data:  data
X-squared = 1.1269, df = 1, p-value = 0.2884

Brain:

data <- data.frame(x= c(6,99), y=c(6,119))
data
chisq.test(data, correct = F)

    Pearson's Chi-squared test

data:  data
X-squared = 0.096461, df = 1, p-value = 0.7561

You can see that the Recurrence p-value doesn't match with what I see in the table. Others, like Brain, Bone, Liver, Lung match with the p-values I calculated.

And for TNBC patients, not a single p-value matches with my calculations. I don't even get Lung and Bone as having significant p-values with my calculations.

I did it for Lung as shown here:

data <- data.frame(x= c(25,143), y=c(9,35))
data
chisq.test(data, correct = F)

    Pearson's Chi-squared test

data:  data
X-squared = 0.8044, df = 1, p-value = 0.3698

But in the paper they mentioned:

We found that high linc-ZNF469-3 expression was significantly correlated to lung metastasis in TNBC patients (Table 1).

Can anyone tell me what could be the wrong here? Is it my calculation wrong? or in the paper?

$\endgroup$
3
  • 1
    $\begingroup$ Ouch. This probably warrants a correction or even retraction. $\endgroup$ Oct 9, 2018 at 11:07
  • $\begingroup$ Exactly. I have another question. Lets say Lung p-value in TNBC patients is significant. High linc-ZNF469-3 has 9 and low linc-ZNF469-3 has 25. How could one tell that high linc-ZNF469-3 expression was significantly correlated to lung metastasis in TNBC patients? $\endgroup$ Oct 9, 2018 at 11:31
  • $\begingroup$ We found that high linc-ZNF469-3 expression was ignificantly correlated to lung metastasis in TNBC patients (Table 1). This conclusion can not be reached. Metastasis across tnbc and non-tnbc require a different analysis e.g. t-test. $\endgroup$ Mar 22, 2020 at 6:55

1 Answer 1

2
$\begingroup$

Your calculations seem right, perhaps there was an error on their side.

I also looked at the number of samples reported, but they use the same amount of samples in each case. Because they are testing whether Linc-ZNF469-3 is different in several tissues, they should have used a multiple test correction for the p-value (although reporting the original values is usually also ok).

To answer your question in the comments. You can't correlate a categorical variable with a continuous variable. Or they did another test or they are not explaining clearly what they did.

$\endgroup$
2
  • $\begingroup$ Thanks for the reply. For my question...I also had a look into some other papers. For eg: Check this table [nature.com/articles/s41598-017-04736-z/tables/1] They say in the paper like this; High RAD51-AS1 expression correlated with advanced FIGO stage, histological subtype, high histological grade, number of tumors and distant metastasis. I'm bit confused. Is it based on their percentage they say RAD51-AS1 expression is correlated with those variables? $\endgroup$ Oct 9, 2018 at 13:53
  • $\begingroup$ @user3351523 They could show a plot of like FIGO stage and the expression of the variable, and see a linear trend towards higher expression on certain stages, but without comparing the mean or the percentage/proportion (and not a correlation) that could be noise or illusions. $\endgroup$
    – llrs
    Oct 9, 2018 at 14:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.