What is the meaning of `batch`
altairwei opened this issue · 6 comments
altairwei commented
Hi,
I had some problems in understanding the batch. In my case, samples were grown under three conditions (control, treatment 1, and treatment 2), and then I obtained samples in three time points (1 to 3 days), with two replicates of each kind. However each sample was grown, sampled and sequenced at a different time. In this sense, each sample is a batch.
The final sample information is shown in the following table:
| SampleID | Time | Treatment | Replicate |
|---|---|---|---|
| S1 | 1 | ctrl | 1 |
| S2 | 1 | ctrl | 2 |
| S3 | 1 | stim1 | 1 |
| S4 | 1 | stim1 | 2 |
| S5 | 1 | stim2 | 1 |
| S6 | 1 | stim2 | 2 |
| ... |
In this case, is there a BATCH?
When I run the following code, which column can be considered as batch ？
res = cna.tl.association(d, #dataset
d.samplem.case, #sample-level attribute of intest (case/control status)
covs=d.samplem[['male']], #covariates to control for (in this case just one)
batches=d.samplem.batch) #batch assignments for each sample so that cna can account for batch effects
yakirr commented
Thanks for your interest in CNA! The "batch" variable is the variable that
CNA will control for the most aggressively (using a permutation test within
each "batch" group). As you anticipate, it doesn't actually have to
literally correspond to an experimental batch. The answer of what variable
should be used as the cna "batch" variable depends on the association test
you're running. Can you tell us which sample attribute you're looking to
associate to your single-cell data?
…On Sun, May 15, 2022 at 10:56 AM Altair Wei ***@***.***> wrote:
Hi,
I had some problems in understanding the batch. In my case, samples were
grown under three conditions (control, treatment 1, and treatment 2), and
then I obtained samples in three time points (1 to 3 days), with two
replicates of each kind. *However each sample was grown, sampled and
sequenced at a different time.* In this sense, each sample is a batch.
The final sample information is shown in the following table:
SampleID Time Treatment Replicate
S1 1 ctrl 1
S2 1 ctrl 2
S3 1 stim1 1
S4 1 stim1 2
S5 1 stim2 1
S6 1 stim2 2
...
In this case, is there a BATCH?
When I run the following code, which column can be considered as batch ？
res = cna.tl.association(d, #dataset
d.samplem.case, #sample-level attribute of intest (case/control status)
covs=d.samplem[['male']], #covariates to control for (in this case just one)
batches=d.samplem.batch) #batch assignments for each sample so that cna can account for batch effects
—
Reply to this email directly, view it on GitHub
<#11>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA4RHON76XWT5PKHHGC2BELVKEGDNANCNFSM5V7GP54A>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
altairwei commented
@yakirr Thank you for your reply! In the case I mentioned above, the sample attributes I am interested in are Time or Treatment.
yakirr commented
Is there a matched design? I.e., do you have a) the same sample at multiple
time points? or b) the same sample with multiple treatments? If so then the
natural batch variable would be sample id so that each sample serves as its
own control. Does that make sense?
…On Mon, May 16, 2022 at 3:21 AM Altair Wei ***@***.***> wrote:
@yakirr <https://github.com/yakirr> Thank you for your reply! In the case
I mentioned above, the sample attributes I am interested in are Time or
Treatment.
—
Reply to this email directly, view it on GitHub
<#11 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA4RHOMOCDBD2EAQCZF5KTTVKHZRBANCNFSM5V7GP54A>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
altairwei commented
@yakirr There are no "same" samples in the biological sense, because they are all grown independently. But in the sense of experimental design, there are two samples (biological replicates) with the same treatment at the same time point, and they can perhaps be called "same sample", right?
yakirr commented
Yes, but you need each batch to contain both case and control samples or
else the permutation test won't work. So you need something broader than
just each pair of replicates. (If there's nothing broader than each pair of
replicates, you can always set the batches=None.)
…On Mon, May 16, 2022 at 3:48 AM Altair Wei ***@***.***> wrote:
@yakirr <https://github.com/yakirr> There are no "same" samples in the
biological sense, because they are all grown independently. But in the
sense of experimental design, there are two samples (biological replicates)
with the same treatment at the same time point, and they can perhaps be
called "same sample", right?
—
Reply to this email directly, view it on GitHub
<#11 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA4RHOOKOWBGSY4H74FYE4DVKH4WNANCNFSM5V7GP54A>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>