Difference between revisions of "Team:Bordeaux/Software"

Line 134: Line 134:
 
</p>
 
</p>
  
<h2>Validating the efficiency of the pipeline results</h2>
+
<img style="width:500px; margin-left:auto; margin-right:auto; display:block" src="https://static.igem.org/mediawiki/2017/4/4a/Bdx-all.png">
 +
 
 +
<h3>3.1. Validating the efficiency of the pipeline results</h2>
 
<p>First of all, to confirm the efficiency of our workflow we decided to look for housekeeping genes behaviors. Among all these genes we have chosen the actin-3. As expected we have been able to locate its junctions in the diagonal area meaning that this particular gene does not have a different alternative splicing between the neuron and muscle. Thus we confirmed the robustness of our pipeline and that allowed us to perform more analysis which are discussed in the following lines.</p>
 
<p>First of all, to confirm the efficiency of our workflow we decided to look for housekeeping genes behaviors. Among all these genes we have chosen the actin-3. As expected we have been able to locate its junctions in the diagonal area meaning that this particular gene does not have a different alternative splicing between the neuron and muscle. Thus we confirmed the robustness of our pipeline and that allowed us to perform more analysis which are discussed in the following lines.</p>
 +
 +
<img style="width:500px; margin-left:auto; margin-right:auto; display:block" src="https://static.igem.org/mediawiki/2017/4/4a/Bdx-all.png">
 +
 +
<h3>3.2. unc-60 splicing investigation</h2>
 +
<p>Since we knew a priori the behavior of unc60, it was an interesting positive control to investigate. We can see on the plot that muscular isoform B and non-muscular isoform A usages behave as expected. Indeed, in the muscle, the usage ratio for UNC-60B is 0.98 versus 0.02 for UNC-60A, a very dichotomic junction usage reflecting the muscle isoform specificity. In contrast, the usages ratios for both isoforms are neighbouring 0.5, which would indicate that both isoforms are used in neuron.</p>
 +
 +
<img style="width:500px; margin-left:auto; margin-right:auto; display:block" src="https://static.igem.org/mediawiki/2017/4/4a/Bdx-all.png">
 +
 +
<h3>3.3. ric-4 splicing investigation</h2>
 +
 +
<p>We had no a priori knowledge about ric-4 but it caught our attention since its behavior is very characteristic of an outlier. Actually its two isoforms are located on the opposite of the diagonal meaning an inversion of spliced forms in comparison with the genes located in the central area. We can see one form very used in the neuron whereas the other one is more used in the muscular tissue.We then investigate the role of ric-4. Thus we found that this gene is involved in the structuration of synapses and their functions.
 +
ric-4 is thought to be related to vesicles trafficking including SNARE vesicles. It is tagged as involved in synapses structuration and function. However SNARE vesicles processes are also found in muscle. Therefore muscle and neuron specific isoforms of these vesicular transport related proteins could exist.</p>
 +
 +
<img style="width:500px; margin-left:auto; margin-right:auto; display:block" src="https://static.igem.org/mediawiki/2017/4/4a/Bdx-all.png">
 +
 +
<h3>3.4. rsr-1 splicing investigation</h2>
 +
 +
<p>rsr-1 was picked up because it presents a splicing pattern very similar to UNC-60. Indeed, rsr-1 isoforms in muscle have poles-apart usage ratios (0.98 vs 0.02) while in neuron this dichotomic usage is quite less pronounced (0.65 vs 0.35). rsr-1 is a homolog of SR160m, a splicing co-activator. It is important for development including normal pharyngeal morphology.
 +
In Ensembl database this gene is featuring only one splice variant. We obtained 7 and 229 read counts for muscular isoforms, and 7 and 13 for the neuron. The few read counts could be due to mapping errors, revealing alternative junctions that are not actually real. This is possible in regions of lower complexity. rsr-1 actually present a low complexity region, long serine and arginine repeats.</p>
  
  
 +
<img style="width:500px; margin-left:auto; margin-right:auto; display:block" src="https://static.igem.org/mediawiki/2017/4/4a/Bdx-all.png">
  
 +
<h2>Discussion and future improvements</h2>
  
 +
<p>
 +
Overall, our observations are supported by decent read counts (junction scores) and known negative and positive controls present expected patterns. However there is still room for amelioration of the pipeline.
 +
</p>
 +
<p>
 +
The area gathering junctions with similar usage ratios is not yet supported by statistical analysis, it is only an arbitrary threshold selected by us. Statistical clustering of junctions is still to be found in order to more robustly separate junctions with similar patterns to those with a significant usage ratio difference.
 +
We are trying to find a method that would be similar to confidence intervals in linear regression analyses.
 +
</p>
 +
<p>
 +
We could rationalise the selection of junction alternatives based on the read count. As well as finding a representation involving this read count.
 +
</p>
  
 
</div>
 
</div>

Revision as of 11:38, 1 November 2017

Wrong