|
⇤ ← Revision 1 as of 2023-07-07 14:05:15
Size: 2810
Comment:
|
Size: 2852
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 29: | Line 29: |
| Note: '''Domain''' is the Uniprot id of SH3 domain containing protein. '''Peptide''' is the Uniprot id of peptide containing protein. '''Start and Stop''' are peptide start and stop positions. '''Sequence''' is the predicted peptide sequence. '''Peptide Score/Protein Score''' is the score of peptide/protein classifier. '''Peptide Count/Protein Count''' is the number of peptide/protein features used for predictions. '''Score''' is the score of combined classifier. | Note: * '''Domain''' is the Uniprot id of SH3 domain containing protein. * '''Peptide''' is the Uniprot id of peptide containing protein. * '''Start and Stop''' are peptide start and stop positions. * '''Sequence''' is the predicted peptide sequence. * '''Peptide Score/Protein Score''' is the score of peptide/protein classifier. * '''Peptide Count/Protein Count''' is the number of peptide/protein features used for predictions. * '''Score''' is the score of combined classifier. |
Predicting in-vivo SH3 domain mediated protein interactions in human.
Shobhit Jain, Ruth Isserlin and Gary Bader
Motivation
SH3 domains mediate many intracellular signaling processes and are critical for cell functioning. These domains bind to proline rich regions which can be identified using high-throughput experimental screens such as phage display. SH3 binding motifs identified by phage display can be used to computationally predict domain mediated protein-protein interactions. The existing landscape of computational approaches for predicting protein interactions is either limited by their inability to predict peptide recognition module mediated interactions or do not consider many known constraints governing these interactions.
Results
A novel method of predicting SH3 domain-peptide mediated protein-protein interactions in humans using phage display data is presented. This method builds upon our previously published work of combining multiple binding site and full length protein features using na{\"i}ve Bayesian models for predicting PRM mediated interactions. In this work, we present a novel algorithm for predicting protein interactions using network topology and show that it outperforms the existing approaches. We have also extended the semi-supervised training regime of multinomial na{\"i}ve Bayesian classifier developed for text classification to Gaussian na{\"i}ve Bayesian models for PPI prediction.
Downloads
Latest Release
Source: DoMo-Pred_human.zip
Predictions
Text file format:
Domain |
Peptide |
Start |
Stop |
Sequence |
Peptide Score |
Peptide Count |
Protein Score |
Protein Count |
Score |
P11710 |
P53861 |
313 |
318 |
RTTSH |
0.96 |
4 |
0.01 |
5 |
0.16 |
P11710 |
P34216 |
236 |
240 |
RTTPL |
0.53 |
4 |
0.98 |
5 |
0.98 |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
Note:
Domain is the Uniprot id of SH3 domain containing protein.
Peptide is the Uniprot id of peptide containing protein.
Start and Stop are peptide start and stop positions.
Sequence is the predicted peptide sequence.
Peptide Score/Protein Score is the score of peptide/protein classifier.
Peptide Count/Protein Count is the number of peptide/protein features used for predictions.
Score is the score of combined classifier.
Supplementary material
PRM_PPI_supplementary_human.pdf
Datasets
Training and test datasets used in manuscript. Datasets_human.zip