Hi David and Chih-Chiang, The recent question was about modifications, which can in principle be treated as different amino acids (which is exactly what they are). For RTCalc and similar models, each modification needs to be present in 15-20 or so distinct peptides for the coefficients to start to converge [and RTCalc needs at least 80 peptides, although 51 (?) are mathematically sufficient].
FYI, I uploaded a Taverna workflow for PeptideProphet pepXML to RTCalc based on a default TPP installation on Windows: http://www.myexperiment.org/workflows/4042.html. This also contains the calls to RTCalc. I'd be curious to know if you can run it and what you think about it! You need R and Rserve (for the statistics and plots) and Taverna workbench to run the workflow, but that should be it. Cheers, Magnus On Monday, 8 December 2014 20:36:33 UTC+1, David Shteynberg wrote: > > In general, RTCalc should be able to handle modified peptides > separately from the unmodified peptides. That said, learning > algorithms such as this perform best when trained with enough data. > If modified peptides are only a small part of the set they may not be > very well characterized by the algorithm. > > -David > > On Sat, Dec 6, 2014 at 12:33 PM, Chih-Chiang Tsou > <[email protected] <javascript:>> wrote: > > Hi David, > > > > Thanks for the reply, does the model take modification into account or > it's > > based on only sequence? And should I remove modified peptides from > training? > > > > Thanks > > Chih-Chiang > > > > On Friday, December 5, 2014 1:25:38 PM UTC-5, David Shteynberg wrote: > >> > >> Hi Chih-Chiang, > >> > >> You are correct, I just checked the code and the PEPXML option has the > >> following comment I left there in the code: > >> > >> if (!pepxmlfile.empty()) { > >> //TODO: DDS Implement rtcalc->parse_pepXML(pepxmlfile); > >> return 0; > >> } > >> > >> So I haven't implemented that feature yet and should remove it from > >> the usage statement. However, you should still be able to get > >> results by pasting a list of peptides and RTs in two columns. The > >> tool has two different training modes. The basic mode uses a linear > >> regression model based on peptide properties, the ANN model uses an > >> artificial neural network based on peptide amino acid positional > >> information and other peptide properties. They are different learning > >> algorithms. > >> > >> -David > >> > >> > >> On Fri, Dec 5, 2014 at 9:36 AM, Chih-Chiang Tsou > >> <[email protected]> wrote: > >> > Hi David, > >> > > >> > I am trying to use rtcalc to train a prediction model directly from > >> > PepXML. > >> > Is the option "PEPXML" for that? it didn't work for me. > >> > Another question is what's the difference between coefficient file > and > >> > ANN > >> > model file? > >> > > >> > Thanks, > >> > Chih-Chiang > >> > > >> > > >> > On Tuesday, April 23, 2013 12:29:18 PM UTC-4, [email protected] > >> > wrote: > >> >> > >> >> Yep - this works fine, thanks! > >> >> > >> >> Cheers, > >> >> > >> >> Magnus > >> >> > >> >> On Wednesday, 17 April 2013 17:49:46 UTC+2, [email protected] > wrote: > >> >>> > >> >>> Hi David, > >> >>> > >> >>> Somehow I missed your reply (which was very quick). I will give it > a > >> >>> try > >> >>> now - thanks! > >> >>> > >> >>> > >> >>> Magnus > >> >>> > >> >>> On Thursday, 28 March 2013 22:28:34 UTC+1, David Shteynberg wrote: > >> >>>> > >> >>>> Thanks for the files you've provided. I was able to find and fix > one > >> >>>> error in the code. TPP revision 6171 from trunk should contain > this > >> >>>> fix. > >> >>>> Also I found an error in your commands. When you train a a Neural > >> >>>> Net with > >> >>>> RTCalc you have to use the ANN= option if you want to then apply > the > >> >>>> trained > >> >>>> model to data. Here are my commands on your files using my new > code: > >> >>>> > >> >>>> > >> >>>> RTCalc TRAIN=list2_predictions_no_neg.txt ANN=list2_ann_DDS.coeff > >> >>>> > >> >>>> > >> >>>> > >> >>>> RTCalc PEPS=peptides.txt ANN=list2_ann_DDS.coeff > >> >>>> > >> >>>> I hope it works for you too! > >> >>>> > >> >>>> -David > >> >>>> > >> >>>> > >> >>>> > >> >>>> On Thu, Mar 28, 2013 at 10:58 AM, [email protected] > >> >>>> <[email protected]> wrote: > >> >>>>> > >> >>>>> OK - I sent the training set and the ANN model by e-mail. > >> >>>>> > >> >>>>> > >> >>>>> On Thursday, 28 March 2013 18:49:41 UTC+1, David Shteynberg > wrote: > >> >>>>>> > >> >>>>>> Hi Magnus, > >> >>>>>> > >> >>>>>> Can you forward me the files you have and the commands you are > >> >>>>>> using > >> >>>>>> and I will debug? > >> >>>>>> > >> >>>>>> Thanks, > >> >>>>>> -David > >> >>>>>> > >> >>>>>> On Thu, Mar 28, 2013 at 10:47 AM, [email protected] > >> >>>>>> <[email protected]> wrote: > >> >>>>>>> > >> >>>>>>> Dear All (especially David), > >> >>>>>>> > >> >>>>>>> I am trying to use the ANN retention time predictor. The > training > >> >>>>>>> runs OK, but when trying to use it I get this error: > >> >>>>>>> > >> >>>>>>> gsl: init_source.c:29: ERROR: vector length n must be positive > >> >>>>>>> integer > >> >>>>>>> Default GSL error handler invoked. > >> >>>>>>> > >> >>>>>>> This application has requested the Runtime to terminate it in > an > >> >>>>>>> unusual way. > >> >>>>>>> Please contact the application's support team for more > >> >>>>>>> information. > >> >>>>>>> > >> >>>>>>> > >> >>>>>>> I used ca. 60,000 different peptides in the training set. How > many > >> >>>>>>> would be needed? What are the things to look out for? > >> >>>>>>> > >> >>>>>>> > >> >>>>>>> Cheers, > >> >>>>>>> > >> >>>>>>> Magnus > >> >>>>>>> > >> >>>>>>> -- > >> >>>>>>> You received this message because you are subscribed to the > Google > >> >>>>>>> Groups "spctools-discuss" group. > >> >>>>>>> To unsubscribe from this group and stop receiving emails from > it, > >> >>>>>>> send an email to [email protected]. > >> >>>>>>> To post to this group, send email to > [email protected]. > >> >>>>>>> > >> >>>>>>> Visit this group at > >> >>>>>>> http://groups.google.com/group/spctools-discuss?hl=en. > >> >>>>>>> For more options, visit > https://groups.google.com/groups/opt_out. > >> >>>>>>> > >> >>>>>>> > >> >>>>>> > >> >>>>>> > >> >>>>> -- > >> >>>>> You received this message because you are subscribed to the > Google > >> >>>>> Groups "spctools-discuss" group. > >> >>>>> To unsubscribe from this group and stop receiving emails from it, > >> >>>>> send > >> >>>>> an email to [email protected]. > >> >>>>> To post to this group, send email to [email protected]. > > >> >>>>> Visit this group at > >> >>>>> http://groups.google.com/group/spctools-discuss?hl=en. > >> >>>>> For more options, visit https://groups.google.com/groups/opt_out. > > >> >>>>> > >> >>>>> > >> >>>> > >> >>>> > >> > -- > >> > You received this message because you are subscribed to the Google > >> > Groups > >> > "spctools-discuss" group. > >> > To unsubscribe from this group and stop receiving emails from it, > send > >> > an > >> > email to [email protected]. > >> > To post to this group, send email to [email protected]. > >> > Visit this group at http://groups.google.com/group/spctools-discuss. > >> > For more options, visit https://groups.google.com/d/optout. > > > > -- > > You received this message because you are subscribed to the Google > Groups > > "spctools-discuss" group. > > To unsubscribe from this group and stop receiving emails from it, send > an > > email to [email protected] <javascript:>. > > To post to this group, send email to [email protected] > <javascript:>. > > Visit this group at http://groups.google.com/group/spctools-discuss. > > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "spctools-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/spctools-discuss. For more options, visit https://groups.google.com/d/optout.
