Extending signalling pathways with protein-interaction networks. Application to apoptosis.

Extending signalling pathways with protein-interaction networks. Application to apoptosis.
Joan Planas-Iglesias, Emre Guney, Javier Garcia-Garcia, Kevin A. Robertson, Sobia Raza, Tom C. Freeman, Peter Ghazal and Baldo Oliva.
OMICS. 2012 May;16(5):245-56. Epub 2012 Mar 2. (PMID: 22385281)

STRING+ expansion network

(Orthology filter at JEV = 1e-20)

Here we present numerical results for STRING+ expansion network in which interactions with JEV > 1e-20 were removed. In the follwing sections we provide the numerals we used to compute the p-Value for our predictions. For the computation of p-Value we consider as candidates the set of nodes in the network with sensitivity higher than 95% and PPV above 90% in the 5-fold cross-validation assay. For each validation set we compute a p-Value describing the likelyhood of obtaining the subset of validated candidates from the total number of proteins in the network according to the sizes of the candidates set and the validation set. Thus, a different p-Value is obtained for each validation set (Keyword, ProMiner).The mean JEV of the candidates and a p-Value for the significance of the differences between the JEV distribution of the candidates and the rest of the nodes in the network is provided as well.

Individual scoring methods

Validation of candidates posed by each scoring method. First column shows the analyzed scoring method. Second column corresponds to the number of nodes that accomplished quality criteria in the requested scoring methods. Columns 3 to 5 show relevant parameters for Keyword validation: the number of validated candidates, the size of the validation set and the p-Value obtained. Columns 6 to 8 contain analogous information for ProMiner validation.

Scoring method	Number of candidates	Keyword validated candidates	Keyword validation size	Keyword validation p-Value	ProMiner validated candidates	ProMiner validation size	ProMiner validation p-Value	JEV mean	JEV differences p-Value
FF	1066	139	584	8.132279e-74	69	297	3.930878e-36	2.821708e-23	2.675464e-19
SC	2620	198	584	2.318265e-63	78	297	3.794338e-19	6.419340e-23	< 1.000000e-323
ALD	2256	173	584	1.723514e-56	79	297	1.676439e-23	8.262728e-23	< 1.000000e-323
iALD	884	134	584	1.441599e-79	50	297	4.605712e-24	6.114888e-23	1.233332e-25

The relevance of sequence similarity between interologs is analyzed in orthology expanded networks by removing those predictions that do not achieve a certain JEV value. Tables corresponding to the orthology filtered versions of this network are available for several -log(JEV) filter thresholds.
To access the tables, follow links below:

none

Go to top of page

Cumulative combinations

Validation of candidates posed by all possible combinations of N scoring methods, where N is shown in the first column. Second column corresponds to the number of nodes that accomplished quality criteria in the requested scoring methods. Columns 3 to 5 show relevant parameters for Keyword validation: the number of validated candidates, the size of the validation set and the p-Value obtained. Columns 6 to 8 contain analogous information for ProMiner validation.

Number of scoring method	Number of candidates	Keyword validated candidates	Keyword validation size	Keyword validation p-Value	ProMiner validated candidates	ProMiner validation size	ProMiner validation p-Value	JEV mean	JEV differences p-Value
4	79	24	584	4.785867e-56	15	297	3.622015e-35	0.000000e+00	3.882790e-69
3	255	53	584	1.172974e-93	34	297	4.376963e-60	1.663464e-24	2.839004e-198
2	1499	145	584	7.461428e-140	65	297	5.299617e-63	2.362012e-23	< 1.000000e-323
1	3326	241	584	7.744366e-152	94	297	1.210206e-59	3.363765e-23	< 1.000000e-323

none

Go to top of page

All Vs all combinations

Validation of candidates posed by each combination of scoring methods. Columns contain relevant parameters for the computation of the p-Value. For each validation method these are: number of candidates posed by the combination of scoring mehtods, number of validated candidates and the p-Value itself. The relevant fields of each validation method are colored in green if the row corresponds to the best combination of scoring methods for the validation method. For best combination of scoring methods rows, links lead to the details of validated protein candidates (under validated candidates columns) or to the details of new protein candidates (under combination of scoring methods column). Signalling-related functions of candidates from best combinations of scoring methods are given in Supplementary table 2.

Combination of scoring methods	Number of candidates	Keyword validated candidates	Keyword validation size	Keyword validation p-Value	ProMiner validated candidates	ProMiner validation size	ProMiner validation p-Value	JEV mean	JEV differences p-Value
SC, ALD	1135	73	584	2.422741e-79	50	297	1.081722e-54	3.091527e-23	< 1.000000e-323
SC, FF	262	45	584	7.332205e-79	36	297	3.897759e-63	2.412077e-24	1.357449e-202
SC, iALD	197	43	584	3.287094e-81	22	297	8.713872e-42	5.580960e-25	1.508032e-144
ALD, FF	218	38	584	6.803658e-70	30	297	2.876369e-55	1.443297e-24	1.219627e-176
ALD, iALD	143	36	584	1.509593e-73	15	297	4.997151e-31	7.660463e-25	1.390985e-107
FF, iALD	291	88	584	1.870561e-151	25	297	6.405927e-43	1.229453e-50	7.250199e-244
SC, ALD, FF	151	30	584	1.848918e-60	29	297	1.883465e-58	2.083700e-24	1.237366e-122
SC, ALD, iALD	112	30	584	7.456043e-65	15	297	1.039827e-32	9.780770e-25	3.010064e-84
SC, FF, iALD	129	35	584	2.761281e-73	20	297	5.512594e-42	0.000000e+00	3.196699e-112
ALD, FF, iALD	100	30	584	1.413332e-66	15	297	1.686778e-33	0.000000e+00	2.975710e-87
SC, ALD, FF, iALD	79	24	584	4.785867e-56	15	297	3.622015e-35	0.000000e+00	3.882790e-69

none

Go to top of page