Targetscan context+ scores, higher or lower

Question

[lishengli19901017]

I'm using Targetscan6.2 to predict miRNA targets. The results coming out really confused me, how to choose among tens of targets of the same miRNA? Targets with higher or lower context+ scores are better? Thank you!

[orzech_mag]

Look at this information, available at MAGIA2 tool site - perhaps, you'll find it interesting:

For each predictor, the stringency of prediction can be tuned by the user.

Different prediction methods uses different strategies to rank predictions: in some cases the more positive the value, the more trustable the prediction, in other cases lower values are better. See the table below to get insights about the range and distribution of scoring values, per species and per predictor. F.i. setting the score lower than -1.4 for microrna_org (H. sapiens) you get no predictions, whereas setting the score to -0.6 you get top 25% trustable predictions.

Picture illustrating values is available here: https://plus.google.com/u/0/101772480484282089828/posts/Co234KAAQVy?pid=6082680808089494978&oid=101772480484282089828

[lishengli19901017]

Thanks, airan. That really helps.

Answer 1

[iraun]

Targets with lowest context+ scores are the most representative ones.

For each predicted target of each miRNA, the sum of the context+ scores for the sites to that miRNA was calculated as the total context+ score. Predicted targets of each miRNA family are sorted by total context+ score. The representative miRNA is the miRNA in its family with the most favorable (lowest) total context+ score.

The context+ score for a specific site is the sum of the contribution of these six features (according to Garcia et al., 2011 paper):

• site-type contribution
• 3' pairing contribution
• local AU contribution
• position contribution
• TA (target site abundance) contribution*
• SPS (seed-pairing stability) contribution*

Here you can find more information: http://www.targetscan.org/vert_61/docs/help.html

Hope it helps.

[1158105901]

thank u iraun that's really helpful