<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class="">The MAKER models will be the final models. Fasta files and features from the raw ab initio gene predictors on the other hand are there for reference purposes only and unless you have a need for them should be ignored. MAKER models are the combination of ab initio gene predictions filtered for best evidence match together with hint based models from the predictors. Basically MAKER took the best models from each separate predictor and created a final consensus gene set. The CDF generator really is for comparison of how evidence match changes between different releases of the genome or for different parameter options (i.e. you are comparing curves between independent MAKER runs and not within a single MAKER run). THE AED CDF curve is interpreted similar to a ROC curve in that shifts up and to the left indicate improved gene models. This is as opposed to using sensitivity and specificity, because those measures require you to already know the correct models in order to generate a comparison. For de-novo annotation that is impossible (if you already had the correct models you wouldn’t be running MAKER), so since such values cannot be generated then AED which used evidence overlap acts as a proxy measurement.<div class=""><div class=""><br class=""></div><div class="">This paper probably gives the overall best example of how AED correlates with model quality (Figures 2 and 3) —> <a href="http://www.biomedcentral.com/1471-2105/12/491" class="">http://www.biomedcentral.com/1471-2105/12/491</a></div><div class=""><br class=""></div><div class="">—Carson</div><div class=""><br class=""></div><div class=""><br class=""></div><div class=""><br class=""></div><div class=""><br class=""><div><blockquote type="cite" class=""><div class="">On Jan 14, 2015, at 1:40 AM, Xabier Vázquez Campos <<a href="mailto:xvazquezc@gmail.com" class="">xvazquezc@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class=""><div class=""><div class=""><div class=""><div class=""><div class="">Hi Maker developers and users,<br class=""><br class=""></div>After quite a bit of time dealing with Maker, I can run it without problems (thank you Carson). However, I have doubts about the evaluation of the best model produced by Maker.<br class=""><br class=""></div>I found the AED_cdf_generator.pl script while searching in the mail list and it is great but, when you use it, what gff files are you comparing? I initially thought that the models to be compared where those from each <i class="">ab initio</i> program, e.g. SNAP vs Augustus, and inside them, the subsequent bootstrap training steps, but unless you run only one each time you run Maker, the XXX.all.gff file will contain data from both predictions. Should I run them individually?<br class=""><br class=""></div>Following the topic, Maker will generate different FASTA files for proteins and transcripts from each program (Maker and each <i class="">ab initio</i> predictor) as well as "non_overlapping" files. Which one(s) do you select to continue with the functional annotation? <br class=""><br class=""></div>Thank you in advance,<br class=""><br class=""></div>Xabier<br clear="all" class=""><div class=""><div class=""><div class=""><div class=""><div class=""><div class=""><br class="">-- <br class=""><div class="gmail_signature">Xabier Vázquez Campos<br class=""><i class="">PhD Candidate</i><br class="">Water Research Centre<br class="">School of Civil and Environmental Engineering<br class="">
The University of New South Wales<br class="">Sydney NSW 2052 AUSTRALIA<br class=""></div>
</div></div></div></div></div></div></div>
_______________________________________________<br class="">maker-devel mailing list<br class=""><a href="mailto:maker-devel@box290.bluehost.com" class="">maker-devel@box290.bluehost.com</a><br class="">http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org<br class=""></div></blockquote></div><br class=""></div></div></body></html>