Dan_YiMi

2014/03/18_Dan_YiMi
由 sufang 在 五, 04/25/2014 - 09:39 發表 Pre-published Dan Robinson RNA-seq

SFL,

I had not heard that news.  I am very doubtful about this.  We find new fusions i.e. new ETS 5’ partners  now and then, but something completely novel detected only by deep sequencing does not smell right.

Functional driving fusions are nearly always expressed at a level that generates dozens to hundreds of fusion reads by our standard sequencing.  Read-throughs, tandem duplications and other subclonal events are expressed at much lower levels, the deeper you sequence the more subclonal events you will see, but they tend to be bystander events. 

I am finishing a composite list of all the recurrent TopHat artifacts we call in our compendium of over 450 samples.  Most of these are not artifacts of the fusion caller, they are real chimeric reads in usually one of four categories. 1. Read-throughs  2. Tandem duplications 3. Overlapping converging or diverging genes 4.Problems with the reference genome / annotation errors. 

When we apply this to most of our TopHat fusion samples >80% of the calls are filtered and we are left with the set of true positives.  I am still finishing this compilation up but I will send it to you when I’m done, you can apply it to your TopHat output.  We used Ensembl annotations in conjunction with TopHat to generate this.  You can call tonight if you have time, we should be home by 10 and are up till 1 usually.

DR