sQTL mapping: how to deal with 0/0 about leafcutter HOT 3 CLOSED

Hi Bosh - I agree that statistically the correct thing to do is to exclude
that observation (that's what the Dirichlet-multinomial would do
implicitly). Yang maybe you can comment on how you did this with
matrixEQTL?

On 28 June 2016 at 11:17, Boxiang Liu [email protected] wrote:

Hi David,

Thanks for making leafcutter available. I am wondering what your strategy
is to deal with 0/0 in sQTL mapping. I have a few thoughts but none are
quite optimal. The first one is to simply through these samples out (for
the particular intron). However, neither matrix eQTL or fastqtl will be
happy with missing data. Another thought is to consider them zero. However
this perhaps does not reflect what is really going on. The final thought is
to maybe use Laplace smoothing, making the proportion 1/N where N is the
total number of introns in the cluster.

Do you have any thoughts on this?

Bosh

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#8, or mute the thread
https://github.com/notifications/unsubscribe/AET_Bhz4BHzEHNHD9BF-OfcveEXqSKqrks5qQWVLgaJpZM4JAYjS
.

David A. Knowles,
Stanford University.
E-mail: [email protected]
Web: http://cs.stanford.edu/~davidknowles/

from leafcutter.

You can try setting the value to the mean/median if not too many are
missing.

Yang

On Tue, Jun 28, 2016 at 11:21 AM, davidaknowles [email protected]
wrote:

Hi Bosh - I agree that statistically the correct thing to do is to exclude
that observation (that's what the Dirichlet-multinomial would do
implicitly). Yang maybe you can comment on how you did this with
matrixEQTL?

On 28 June 2016 at 11:17, Boxiang Liu [email protected] wrote:

Hi David,

Thanks for making leafcutter available. I am wondering what your strategy
is to deal with 0/0 in sQTL mapping. I have a few thoughts but none are
quite optimal. The first one is to simply through these samples out (for
the particular intron). However, neither matrix eQTL or fastqtl will be
happy with missing data. Another thought is to consider them zero.
However
this perhaps does not reflect what is really going on. The final thought
is
to maybe use Laplace smoothing, making the proportion 1/N where N is the
total number of introns in the cluster.

Do you have any thoughts on this?

Bosh

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#8, or mute the
thread
<
https://github.com/notifications/unsubscribe/AET_Bhz4BHzEHNHD9BF-OfcveEXqSKqrks5qQWVLgaJpZM4JAYjS

.

David A. Knowles,
Stanford University.
E-mail: [email protected]
Web: http://cs.stanford.edu/~davidknowles/

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#8 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AE1pia39VrLMfZbYh0rDZR-BXU_Q_TlJks5qQWYbgaJpZM4JAYjS
.

from leafcutter.

That makes sense. Thank you!

from leafcutter.