Daniel:Notebook/Haplotyping/SISSORPipeline/2014-12-5: Difference between revisions

From ZhangLabWiki
Jump to navigation Jump to search
>Djacobse
(Created page with "=Segmentation Analysis Pipeline (Started 12/2/2014)= Back to Notebook ...")
 
>Djacobse
 
Line 6: Line 6:


<gallery widths=500px heights=500px>
<gallery widths=500px heights=500px>
File:
File:Negative_Binomial_Fit.jpg|Negative Binomial distribution fit
File:
File:Gamma_Fit_Chamber_1.jpg|Gamma distribution fit
</gallery>
</gallery>



Latest revision as of 00:10, 9 December 2014

Segmentation Analysis Pipeline (Started 12/2/2014)[edit]

Back to Notebook

On 12/3, I identified that the cutoff of 20 reads used in the pipeline is actually pretty arbitrary, and is based off the expected number of reads per bin based on the amount of sequencing we do. As a result, I analyzed the read count distribution from Chamber 1 of PGP1-22 (arbitrarily chosen). I loaded the readcount data into MATLAB, and attempted to fit the distribution using the fitdist command. The code is here The two most closely fitting distributions are the Negative Binomial and the Gamma, which makes sense based on the underlying assumptions of the probability distributions. For reference, I have included the CDFs of the two distributions below.

Conclusion[edit]

After talking with Dr. Zhang, we have decided to keep the 20 for now, and that I may look in to the cutoff later, if it becomes a bigger problem.