search

gdario / deep_ppi

A TensorFlow/Python3 refactoring of the Deep PPI project
4 stars 3 forks source link

Need for protein feature extraction script #1

Open SumeetTiwari07 opened 6 years ago

SumeetTiwari07 commented 6 years ago

Thanks for providing the updated scripts for keras 2.0. It's really a great help if the script for the feature extraction of proteins has also been provided. Show that user can make their own model and predict the interaction of their desired protein. Because without feature extraction we cannot use these scripts.

gdario commented 6 years ago

Thanks for pointing this out! Admittedly I haven't checked the code of recent. I have added a link to the README file that links back to the original data and code. That code is still based on Keras 1.2, but it should be straightforward to refactor. I will definitely update the code following your suggestion, but I'm afraid I won't be able to do it for at least one week.

SumeetTiwari07 commented 6 years ago

Thank for the quick response. But i have seen all the codes and data provided by the authors along with this paper but they even did not provided that script. I hope they would have done that. That would be a great help to everyone. But really thanks for your efforts as u made it keras 2.0. That will be great if you could able to get that script from the authors.

On Dec 8, 2017 8:30 PM, "gdario" notifications@github.com wrote:

Thanks for pointing this out! Admittedly I haven't checked the code of recent. I have added a link to the README file that links back to the original data and code. That code is still based on Keras 1.2, but it should be straightforward to refactor. I will definitely update the code following your suggestion, but I'm afraid I won't be able to do it for at least one week.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gdario/deep_ppi/issues/1#issuecomment-350351206, or mute the thread https://github.com/notifications/unsubscribe-auth/Agx_PKJUdbHTmZ0n93ijVWzB2Jd10MUeks5s-Y5KgaJpZM4Q7FOF .

gdario commented 6 years ago

I see, sorry for not getting your point! I haven't re-read the paper but I don't think the authors mention anywhere which package they used to produce the protein fingerprints. For a different project, I used the protr package. I used the same predictors they mention, but I didn't end up with the same number of predictors they have. That's not surprising because for a number of the predictors you must specify some extra arguments, and I had to select different values from theirs. However, I wouldn't expect the results to be significantly different, although it's difficult to say. Honestly I've never contacted the authors directly, so I think you have the same chance I have of getting the code from them. I'll try when I'm back (I'm flying back from NIPS 2017 tonight), and if I have any luck, I'll let you know.

SumeetTiwari07 commented 6 years ago

Thanks for providing the link to the package protr. I will look at the package and will try to sort out with this. But yeah i have wrote to them 2-3 times but they have not responded yet. But its fine you carry on with your trip.

On Sun, Dec 10, 2017 at 10:46 PM, gdario notifications@github.com wrote:

I see, sorry for not getting your point! I haven't re-read the paper but I don't think the authors mention anywhere which package they used to produce the protein fingerprints. For a different project, I used the protr https://cran.r-project.org/web/packages/protr/index.html package. I used the same predictors they mention, but I didn't end up with the same number of predictors they have. That's not surprising because for a number of the predictors you must specify some extra arguments, and I had to select different values from theirs. However, I wouldn't expect the results to be significantly different, although it's difficult to say. Honestly I've never contacted the authors directly, so I think you have the same chance I have of getting the code from them. I'll try when I'm back (I'm flying back from NIPS 2017 tonight), and if I have any luck, I'll let you know.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gdario/deep_ppi/issues/1#issuecomment-350584105, or mute the thread https://github.com/notifications/unsubscribe-auth/Agx_PBEINcAU5N68jxj__kT5oCQsP7jPks5s_FEbgaJpZM4Q7FOF .

-- Regards, Sumeet Kumar Tiwari, Phd Student, NG 1 Mikrobielle Genomik, Robert Koch-Institut, Berlin, Germany.

SumeetTiwari07 commented 6 years ago

Hi Just for the information this the online server that authors has used to extract the features.After an extensive search I have found it show thought to share with you may it will be useful to u as well. The web server name is PROFEAT. I have compared the results it was totally same. This the link to that server http://bidd2.nus.edu.sg/cgi-bin/prof2015/protein/profnew.cgi

On Mon, Dec 11, 2017 at 10:29 AM, SUMEET TIWARI sumeet.kumartt@gmail.com wrote:

Thanks for providing the link to the package protr. I will look at the package and will try to sort out with this. But yeah i have wrote to them 2-3 times but they have not responded yet. But its fine you carry on with your trip.

On Sun, Dec 10, 2017 at 10:46 PM, gdario notifications@github.com wrote:

I see, sorry for not getting your point! I haven't re-read the paper but I don't think the authors mention anywhere which package they used to produce the protein fingerprints. For a different project, I used the protr https://cran.r-project.org/web/packages/protr/index.html package. I used the same predictors they mention, but I didn't end up with the same number of predictors they have. That's not surprising because for a number of the predictors you must specify some extra arguments, and I had to select different values from theirs. However, I wouldn't expect the results to be significantly different, although it's difficult to say. Honestly I've never contacted the authors directly, so I think you have the same chance I have of getting the code from them. I'll try when I'm back (I'm flying back from NIPS 2017 tonight), and if I have any luck, I'll let you know.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gdario/deep_ppi/issues/1#issuecomment-350584105, or mute the thread https://github.com/notifications/unsubscribe-auth/Agx_PBEINcAU5N68jxj__kT5oCQsP7jPks5s_FEbgaJpZM4Q7FOF .

-- Regards, Sumeet Kumar Tiwari, Phd Student, NG 1 Mikrobielle Genomik, Robert Koch-Institut, Berlin, Germany.

-- Regards, Sumeet Kumar Tiwari, Phd Student, NG 1 Mikrobielle Genomik, Robert Koch-Institut, Berlin, Germany.

gdario commented 6 years ago

Thanks a lot for the heads-up! That's really useful! I'm curious to see whether I get the same values I get from protr. For most predictors this should be the case.

SumeetTiwari07 commented 6 years ago

Yeah that will be fine. Let me know the results too. As you have modified the codes so i think you understood the complete code well. As i am bit new to deep learning field and don't know much python. I would also like to know how we can build our own model using those code for prediction by using those code. Because i tried but didn't obtain the same model as the author has given in the data.

Thank you,

On Fri, Dec 15, 2017 at 11:40 AM, gdario notifications@github.com wrote:

Thanks a lot for the heads-up! That's really useful! I'm curious to see whether I get the same values I get from protr. For most predictors this should be the case.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gdario/deep_ppi/issues/1#issuecomment-351973512, or mute the thread https://github.com/notifications/unsubscribe-auth/Agx_PBNKXK0o5mmo8tMUASMeaK_-v8Yxks5tAkysgaJpZM4Q7FOF .

-- Regards, Sumeet Kumar Tiwari, Phd Student, NG 1 Mikrobielle Genomik, Robert Koch-Institut, Berlin, Germany.