Open annashcherbina opened 7 years ago
One idea we discussed earlier was removing key drivers of PC1, if PC1 did not correlate to DMSO treatment or timepoint, but it looks like that is not the case -- PC1 splits between DMSO treated and control in the rlogTransform(counts) space:
That's great. This makes sense to me. Can discuss in person tomorrow briefly.
Anshul
On Apr 27, 2017 5:11 PM, "annashcherbina" notifications@github.com wrote:
Re-running PCA on rlogTransform(counts) suggests the absence of surrogate variables.
One idea we discussed earlier was removing key drivers of PC1, if PC1 did not correlate to DMSO treatment or timepoint, but it looks like that is not the case -- PC1 splits between DMSO treated and control in the rlogTransform(counts) space:
[image: rlogtransformcounts_varplot] https://cloud.githubusercontent.com/assets/5261545/25509161/ca46fa38-2b6b-11e7-804a-67abf7957fc9.png [image: rlogtransform_counts_pca] https://cloud.githubusercontent.com/assets/5261545/25509165/cf3f72ae-2b6b-11e7-9168-fece0110ceb7.png Running PCA on asinh(tpm):
[image: asinh_tpm_varplot] https://cloud.githubusercontent.com/assets/5261545/25509179/e7984aa6-2b6b-11e7-9701-f387492eedfe.png [image: asinh_tpm_pca] https://cloud.githubusercontent.com/assets/5261545/25509180/e9b65e7c-2b6b-11e7-8868-4cbadda6d9ba.png and with removal of the surrogate variables:
[image: asinh_tpm_minus_sva_varplot] https://cloud.githubusercontent.com/assets/5261545/25509266/86a5b1b0-2b6c-11e7-8971-4f7078adbd00.png [image: asinh_tpm_minus_sva_pca] https://cloud.githubusercontent.com/assets/5261545/25509268/89d68378-2b6c-11e7-9558-3c4781cc0b8a.png
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kundajelab/DMSO/issues/6#issuecomment-297872898, or mute the thread https://github.com/notifications/unsubscribe-auth/AAI7ESMd_okdYXxfekBDrcOdCB91Ug4Kks5r0S6tgaJpZM4NCY0I .