The AFM-PISA-classifier package uses AlphaFold-Multimer (AFM) predictions of the human reference set of positive and random protein interactions (hsPRS-v2, hsRRS-v2) from ‘Choi, S.G., Olivet, J., Cassonnet, P. et al. Nat Communications (2019)’ as training set to predict the likelihood of AFM predictions of interest to be true-positive interactions. For this, the AFM predicted structures are analyzed using the PDBePISA tool for the exploration of macromolecular interfaces. As training features, the PAE values from the AFM predictions and interface areas retrieved by PDBePISA are used.
For some plots XQuartz is required, which can be downloaded from XQuartz. You can install the development version of AFM-PISA-classifier from GitHub with:
# Install dependencies
install.packages("BiocManager")
BiocManager::install("ComplexHeatmap")
BiocManager::install("Biostrings")
BiocManager::install("BiocStyle")
devtools::install_github("philipptrepte/binary-PPI-classifier")
# Install AFMpisa
devtools::install_github("philipptrepte/AFM-Pisa-classifier")Predict protein complexes with AlphaFold-Multimer using for example AlphaFold or ColabFold. Results for each predicted model will include ‘.json’ files containing the PAE values.
Use PDBePISA or PisaPy for batch analysis of AFM predictions to extract information on macromolecular interfaces. Results will include ‘[…]interfacesummary0.xml’ and ‘[…]interfacetable.xml’ files. Among others, the interfacesummary contains information on the ‘Interface Area’ and ‘Number of Interface Residues’, while the interfacetable contains information on the ‘Total Surface Area’, the ‘Solvation Free Energy’, ‘H-Bonds’, ‘Saltbridges’, ‘Disulfide Bonds’ and ‘Complexation Significance Score (CSS)’.
Evaluation of AFM predicted structural models, requires the
binaryPPIclassifier
package. To evaluate the predicted models, you can either train a
multi-adaptive support vector machine learning (maSVM) algorithm on your
own training data, or use the provided models, which can be found in the
/data/maSVM_models/ folder. The provided models were generated by
training the maSVM algorithm on AFM predicted structural models from 51
known interactions and 67 random protein pairs not known to interact.
The function import.afm() will import all .json files in the specified
directory dir where your save the results from the AFM predictions.
PAE, pLLDT and PTM values will be extracted and stored in a list.
In the following, we provide an example for the AFM predicted structural models for the NSP10-NSP16 and NSP10-NSP14 complexes.
YOUR_AFM_RESULTS <- import.afm(dir = "data/AFM_json/")
#> read AFM `config.json` files: 1 of 3
#> read AFM `*scores.json` files: 1 of 15
#> extract ptm, plldt and pae values: 1 of 15
#> Done.
#> extract ptm, plldt and pae values: 1 of 15
#> Done.
#> read `*.a3m` files: 1 of 3
#> read `*.a3m` files: 3 of 3The list will contain a data frame providing you with an overview of the predicted protein complexes:
head(YOUR_AFM_RESULTS$protein)
#> A_length B_length A_protein B_protein
#> 1 75 222 E M
#> 2 139 298 NSP10 NSP16
#> 3 527 139 NSP14 NSP10It also contains a vector with pLLDT values for every amino acid:
YOUR_AFM_RESULTS$plldt$`NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_1_model_3_scores.json`
#> [1] 19.07 22.83 28.73 31.85 34.60 41.77 45.49 59.83 68.14 85.21 84.39 85.99
#> [13] 92.89 94.01 91.93 92.82 96.17 94.03 94.74 95.75 94.84 96.28 97.28 97.10
#> [25] 97.11 97.60 97.92 96.96 96.12 95.75 95.66 95.89 95.46 94.40 93.67 94.37
#> [37] 96.53 96.69 94.67 94.92 98.07 98.71 98.78 98.75 98.47 96.32 96.04 96.50
#> [49] 95.86 96.01 95.48 96.37 97.41 98.22 98.32 97.83 96.48 96.63 97.98 97.36
#> [61] 97.92 97.28 95.65 96.37 98.02 98.34 98.78 98.89 98.82 98.58 97.53 98.18
#> [73] 98.50 97.56 97.48 95.81 95.25 95.54 94.31 90.71 92.39 91.32 92.79 89.33
#> [85] 84.07 79.61 78.16 77.42 84.43 93.18 92.44 97.04 95.63 96.24 97.78 98.66
#> [97] 98.84 98.73 98.53 97.98 97.17 96.49 97.24 96.73 96.02 97.60 98.44 98.36
#> [109] 97.99 98.23 97.79 96.89 95.63 96.47 94.54 96.18 96.08 93.59 93.93 95.10
#> [121] 96.29 97.31 97.23 95.24 92.36 93.59 89.21 91.61 87.72 87.09 70.62 60.53
#> [133] 55.81 51.93 49.06 45.46 43.69 38.55 28.60 44.57 64.90 75.27 83.50 90.52
#> [145] 94.97 97.80 97.73 97.98 98.25 98.54 97.77 96.67 97.23 97.80 96.61 95.69
#> [157] 95.30 97.26 96.28 95.23 95.25 93.43 93.13 94.21 94.91 95.40 92.35 92.97
#> [169] 92.60 91.83 93.37 89.20 88.00 92.76 95.41 94.13 93.28 94.25 97.37 96.99
#> [181] 96.43 97.43 98.61 98.45 98.41 98.84 98.82 98.63 98.87 98.87 98.55 98.55
#> [193] 98.70 98.25 97.42 97.72 96.88 98.12 98.71 98.67 97.85 95.83 94.56 97.76
#> [205] 98.37 98.74 98.40 98.68 98.37 97.50 97.64 95.57 92.76 88.08 90.49 92.63
#> [217] 96.66 97.91 98.56 98.72 98.75 98.76 98.87 98.90 98.47 98.41 98.69 98.55
#> [229] 97.43 94.63 93.68 96.93 97.12 97.91 97.36 97.37 96.64 95.70 91.54 88.52
#> [241] 89.63 93.12 94.93 96.81 97.56 97.36 95.10 93.00 94.35 94.08 93.19 90.92
#> [253] 91.91 92.52 91.51 90.92 91.39 90.77 91.49 90.62 94.98 97.57 98.43 98.75
#> [265] 98.94 98.92 98.90 98.77 98.01 96.40 88.48 82.82 75.17 74.94 67.76 64.45
#> [277] 62.63 61.50 60.46 66.06 79.78 87.93 89.37 90.76 90.37 89.29 87.54 94.42
#> [289] 96.56 96.24 96.25 97.79 98.39 98.06 98.01 98.64 98.29 97.38 97.96 98.73
#> [301] 98.68 98.70 98.81 98.90 98.97 98.97 98.97 98.92 98.79 97.95 96.55 94.92
#> [313] 93.89 94.79 96.34 97.30 98.26 98.37 98.76 98.68 98.61 98.76 98.78 98.59
#> [325] 98.73 98.83 98.60 98.16 98.64 98.72 98.84 98.83 98.33 98.10 97.24 97.30
#> [337] 96.56 96.71 96.56 95.07 95.37 97.24 98.57 98.89 98.90 98.94 98.93 98.78
#> [349] 98.83 98.86 98.24 95.91 94.66 95.44 95.49 95.52 96.99 98.02 98.13 98.22
#> [361] 97.79 98.28 98.61 98.43 98.43 98.71 98.45 97.86 97.97 98.26 97.69 96.93
#> [373] 96.23 93.63 90.77 93.19 91.85 95.51 95.59 96.13 95.29 96.45 97.17 94.23
#> [385] 91.15 95.20 92.10 94.26 97.68 97.71 98.21 95.62 88.42 87.96 91.60 97.21
#> [397] 97.44 97.73 97.27 96.24 95.14 92.31 88.39 92.03 95.07 95.45 96.19 96.34
#> [409] 97.36 97.67 97.76 97.99 98.40 98.51 98.38 97.99 98.41 98.44 98.61 98.58
#> [421] 98.53 96.87 93.61 86.76 84.11 82.29 83.25 79.39 79.53 76.29 78.37 60.20
#> [433] 65.61 59.12 54.24 64.36 57.44It also contains a matrix with PAE values for every amino acid pair. Only the first 20 amino acids are shown:
YOUR_AFM_RESULTS$pae$`NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_1_model_3_scores.json`[1:20,1:20]
#> [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12]
#> [1,] 0.75 1.49 4.12 5.94 7.65 11.52 13.78 15.38 17.86 17.29 19.35 17.79
#> [2,] 1.96 0.75 1.25 3.80 6.29 8.62 10.77 13.07 14.73 15.19 17.10 15.42
#> [3,] 4.49 2.26 0.75 1.38 4.45 6.70 7.83 9.96 11.75 13.09 14.27 13.40
#> [4,] 6.38 3.91 1.96 0.75 1.46 4.70 6.99 8.18 10.84 12.76 13.07 12.52
#> [5,] 7.80 5.67 4.15 2.03 0.75 1.39 4.81 6.53 8.88 10.29 10.76 10.47
#> [6,] 10.26 8.07 6.84 4.86 2.14 0.75 1.40 4.20 6.79 7.18 7.85 8.31
#> [7,] 13.47 10.40 8.23 6.64 4.72 2.00 0.75 1.20 4.28 5.53 6.21 6.29
#> [8,] 15.51 13.27 10.44 8.62 7.50 3.97 1.59 0.75 1.02 3.08 4.06 4.18
#> [9,] 17.18 15.48 14.16 11.67 9.81 6.78 3.65 1.57 0.75 1.98 2.22 2.81
#> [10,] 20.81 18.85 16.31 13.97 11.05 9.37 5.80 4.16 1.87 0.75 0.90 1.96
#> [11,] 20.19 17.87 13.75 12.95 10.99 7.78 6.51 5.57 3.60 1.17 0.75 0.84
#> [12,] 23.05 19.03 17.56 15.17 13.45 10.08 8.81 6.31 4.90 2.64 1.21 0.75
#> [13,] 24.92 22.75 20.58 17.90 14.31 11.22 9.42 6.40 4.21 2.84 2.28 0.86
#> [14,] 24.33 23.00 20.28 17.76 14.75 11.31 9.97 8.56 4.65 2.77 2.57 1.76
#> [15,] 25.13 23.17 19.51 17.94 17.20 13.02 12.20 9.73 6.56 3.25 2.18 1.79
#> [16,] 26.31 24.78 22.70 20.80 18.53 14.84 14.65 10.92 7.13 4.17 2.40 1.66
#> [17,] 27.25 25.64 24.26 22.79 19.18 14.70 13.89 11.62 5.79 4.91 3.64 2.10
#> [18,] 27.10 25.45 23.53 22.32 20.41 15.39 15.20 13.87 7.41 5.60 3.74 2.76
#> [19,] 27.85 26.49 24.40 23.60 23.10 19.83 17.79 16.49 10.69 5.46 4.34 3.25
#> [20,] 28.50 27.20 25.26 24.14 22.97 18.94 17.93 14.96 8.76 6.12 4.15 2.66
#> [,13] [,14] [,15] [,16] [,17] [,18] [,19] [,20]
#> [1,] 17.31 17.72 17.64 17.67 17.34 18.04 18.17 18.77
#> [2,] 15.57 15.87 15.48 15.56 15.33 16.07 15.99 16.87
#> [3,] 13.56 13.75 13.36 13.38 13.44 14.20 14.09 14.70
#> [4,] 12.34 12.60 12.26 12.04 12.28 12.77 12.92 13.39
#> [5,] 10.19 10.57 10.31 9.99 10.11 10.66 10.93 11.07
#> [6,] 8.16 8.94 8.16 7.98 7.81 8.21 8.95 8.79
#> [7,] 6.13 6.49 6.58 6.11 6.23 6.43 6.97 6.21
#> [8,] 3.99 4.25 4.02 3.89 4.06 4.46 4.79 4.47
#> [9,] 2.85 2.77 2.57 2.86 2.97 3.38 3.89 3.52
#> [10,] 1.63 1.86 1.64 1.84 1.94 2.25 2.81 2.96
#> [11,] 1.39 1.50 1.50 1.84 1.63 1.99 2.66 2.50
#> [12,] 0.83 1.02 1.51 1.28 1.39 1.79 2.57 2.17
#> [13,] 0.75 0.76 1.01 1.09 0.98 1.22 1.94 1.73
#> [14,] 0.79 0.75 0.77 0.99 0.98 1.08 1.57 1.56
#> [15,] 1.07 0.78 0.75 0.78 0.90 1.13 1.50 1.56
#> [16,] 1.16 0.95 0.82 0.75 0.77 1.01 1.60 1.41
#> [17,] 1.14 1.01 1.03 0.79 0.75 0.77 1.17 1.19
#> [18,] 1.32 1.08 1.14 1.03 0.78 0.75 0.80 1.10
#> [19,] 1.82 1.41 1.44 1.27 1.03 0.78 0.75 0.80
#> [20,] 1.72 1.49 1.46 1.27 1.07 1.04 0.86 0.75It also contains a data frame on the PTM and max PAE values:
YOUR_AFM_RESULTS$ptm
#> file
#> 1 E-M_53cd1.result/E_M_53cd1_unrelaxed_rank_1_model_2_scores.json
#> 2 E-M_53cd1.result/E_M_53cd1_unrelaxed_rank_2_model_4_scores.json
#> 3 E-M_53cd1.result/E_M_53cd1_unrelaxed_rank_3_model_1_scores.json
#> 4 E-M_53cd1.result/E_M_53cd1_unrelaxed_rank_4_model_3_scores.json
#> 5 E-M_53cd1.result/E_M_53cd1_unrelaxed_rank_5_model_5_scores.json
#> 6 NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_1_model_3_scores.json
#> 7 NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_2_model_2_scores.json
#> 8 NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_3_model_1_scores.json
#> 9 NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_4_model_5_scores.json
#> 10 NSP10-NSP16_3cd38.result/NSP10_NSP16_3cd38_unrelaxed_rank_5_model_4_scores.json
#> 11 NSP14-NSP10_c009a.result/NSP14_NSP10_c009a_unrelaxed_rank_1_model_1_scores.json
#> 12 NSP14-NSP10_c009a.result/NSP14_NSP10_c009a_unrelaxed_rank_2_model_2_scores.json
#> 13 NSP14-NSP10_c009a.result/NSP14_NSP10_c009a_unrelaxed_rank_3_model_5_scores.json
#> 14 NSP14-NSP10_c009a.result/NSP14_NSP10_c009a_unrelaxed_rank_4_model_4_scores.json
#> 15 NSP14-NSP10_c009a.result/NSP14_NSP10_c009a_unrelaxed_rank_5_model_3_scores.json
#> ptm max_pae
#> 1 0.41 31.75
#> 2 0.41 31.75
#> 3 0.37 31.75
#> 4 0.36 31.75
#> 5 0.42 31.75
#> 6 0.92 31.75
#> 7 0.92 31.75
#> 8 0.92 31.75
#> 9 0.92 31.75
#> 10 0.92 31.75
#> 11 0.90 31.75
#> 12 0.89 31.75
#> 13 0.89 31.75
#> 14 0.88 31.75
#> 15 0.88 31.75And finally the number of models predicted by AFM for each protein pair:
YOUR_AFM_RESULTS$num_models
#> [1] 5 5 5The function import.pisa() will import from all subdirectoris under
the specified directory dir the interfacesummary0.xml and
interfacetable.xml files that you saved from the PDBePISA results.
Information on ‘Interface Area’, ‘Number of Interface Residues’, ‘Total
Surface Area’, ‘Solvation Free Energy’, ‘H-Bonds’, ‘Saltbridges’,
‘Disulfide Bonds’ and ‘Complexation Significance Score (CSS)’ will be
extracted and stored in a data frame.
In the following, we provide an example on the information extracted from the PDBePISA analyzed AFM predicted structural models of the E-M, NSP10-NSP16 and NSP10-NSP14 complexes.
YOUR_PISA_INTERFACE <- import.pisa(dir = "data/PDBePISA_xml/")
YOUR_PISA_INTERFACE
#> # A tibble: 15 × 17
#> complex model rank surfaceAreaA surfaceAreaB interfaceAreaA interfaceAreaB
#> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 E_M 2 1 7810. 17071. 1333. 1306.
#> 2 E_M 4 2 7917. 16348 1021. 913.
#> 3 E_M 1 3 7629. 17173. 1053. 918.
#> 4 E_M 3 4 7878. 17654. 1001. 1068.
#> 5 E_M 5 5 7641. 16512. 810. 764.
#> 6 NSP10_NS… 3 1 8915. 14330. 986. 923.
#> 7 NSP10_NS… 2 2 8919. 14331 974. 902.
#> 8 NSP10_NS… 1 3 8706. 14215. 1162. 1094.
#> 9 NSP10_NS… 5 4 8750. 14314. 1263. 1193.
#> 10 NSP10_NS… 4 5 8766. 14186. 1202. 1110.
#> 11 NSP14_NS… 1 1 26445. 8903. 2176. 2364.
#> 12 NSP14_NS… 2 2 26911. 8903. 2239. 2378.
#> 13 NSP14_NS… 5 3 26758. 8890. 2178. 2313.
#> 14 NSP14_NS… 4 4 26800. 8901. 2180. 2406.
#> 15 NSP14_NS… 3 5 27026. 8867. 2131. 2286.
#> # ℹ 10 more variables: interfaceResiduesA <dbl>, interfaceResiduesB <dbl>,
#> # deltaG <dbl>, pvalue <dbl>, hbonds <dbl>, saltbridges <dbl>,
#> # disulfide <dbl>, css <dbl>, interfaceArea <dbl>, surfaceArea <dbl>This function plots the pLLDT values for each amino acid of an AFM
predicted structural model. You need to specify which complex to plot
afm_complex = "NSP10-NSP16" and which rank afm_rank = 1 or
afm_rank = "all".
plot_grid(
plldt.lineplot(import_afm = YOUR_AFM_RESULTS, afm_complex = "E-M", afm_rank = 'all'),
plldt.lineplot(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP10-NSP16", afm_rank = 'all'),
plldt.lineplot(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP14-NSP10", afm_rank = 'all'),
ncol = 1
)
The function pae.heatmap() lets you visualize the intra- and inter-residue PAE values from your AFM predicted structural models. You can specify which AFM model to plot by providing the rank, and to only plot the kmeans clustered interface region for which you can set a pLLDT cutoff that defines which amino acids to include during kmeans clustering. When clustering the interface region, a barplot will be plotted, showing the average PAE values from the resulting eight clusters.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "E-M", afm_rank = 1, interface_cluster = FALSE)
Note that we used here a pLLDT of 0 `plldt = 0’ as cutoff.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "E-M", afm_rank = 1, interface_cluster = TRUE, plldt = 0)
#> pae mean of stats::kmeans-clustering: 1 of 15
#> pae mean of stats::kmeans-clustering: 10 of 15
#> Done.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP10-NSP16", afm_rank = 1, interface_cluster = FALSE)
Note that we used here a pLLDT of 50 `plldt = 50’ as cutoff.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP10-NSP16", afm_rank = 1, interface_cluster = TRUE, plldt = 50)
#> pae mean of stats::kmeans-clustering: 1 of 15
#> pae mean of stats::kmeans-clustering: 10 of 15
#> Done.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP14-NSP10", afm_rank = 1, interface_cluster = FALSE)
Note that we used here a pLLDT of 50 `plldt = 50’ as cutoff.
pae.heatmap(import_afm = YOUR_AFM_RESULTS, afm_complex = "NSP14-NSP10", afm_rank = 1, interface_cluster = TRUE, plldt = 50)
#> pae mean of stats::kmeans-clustering: 1 of 15
#> pae mean of stats::kmeans-clustering: 10 of 15
#> Done.
This function performs kmeans clustering to identify amino acid clusters
with the lowest average PAE. This is equivalent to the results from the
pae.heatmap() function with the parameter interface_cluster = TRUE.
The function takes the resulting list from import.afm() as input and
produces a data frame as output, which stores for the resulting eight
clusters (AB cluster 1-4 and BA cluster 1-4) information on their median
and mean PAE values as well as the cluster size as
number of residues protein A * number of residues protein B. Note that
we used a pLLDT cutoff of >50 for all complexes now plldt = 50.
YOUR_INTERFACE <- pae.interface(import_afm = YOUR_AFM_RESULTS, plldt = 50)
#> pae mean of stats::kmeans-clustering: 1 of 15
#> pae mean of stats::kmeans-clustering: 10 of 15
#> Done.colnames(YOUR_INTERFACE)
#> [1] "A_length" "B_length"
#> [3] "A_protein" "B_protein"
#> [5] "file" "complex"
#> [7] "model" "rank"
#> [9] "interAB.cluster1.median" "interAB.cluster1.mean"
#> [11] "interAB.cluster1.size" "interAB.cluster2.median"
#> [13] "interAB.cluster2.mean" "interAB.cluster2.size"
#> [15] "interAB.cluster3.median" "interAB.cluster3.mean"
#> [17] "interAB.cluster3.size" "interAB.cluster4.median"
#> [19] "interAB.cluster4.mean" "interAB.cluster4.size"
#> [21] "interBA.cluster1.median" "interBA.cluster1.mean"
#> [23] "interBA.cluster1.size" "interBA.cluster2.median"
#> [25] "interBA.cluster2.mean" "interBA.cluster2.size"
#> [27] "interBA.cluster3.median" "interBA.cluster3.mean"
#> [29] "interBA.cluster3.size" "interBA.cluster4.median"
#> [31] "interBA.cluster4.mean" "interBA.cluster4.size"
#> [33] "pae" "interaction"
YOUR_INTERFACE %>% dplyr::select(A_length, B_length, A_protein, B_protein, complex, model, rank, pae)
#> # A tibble: 15 × 8
#> A_length B_length A_protein B_protein complex model rank pae
#> <dbl> <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl>
#> 1 75 222 E M E_M 2 1 15.3
#> 2 75 222 E M E_M 4 2 16.3
#> 3 75 222 E M E_M 1 3 17.8
#> 4 75 222 E M E_M 3 4 15.2
#> 5 75 222 E M E_M 5 5 19.0
#> 6 139 298 NSP10 NSP16 NSP10_NSP16 3 1 3.77
#> 7 139 298 NSP10 NSP16 NSP10_NSP16 2 2 3.30
#> 8 139 298 NSP10 NSP16 NSP10_NSP16 1 3 3.87
#> 9 139 298 NSP10 NSP16 NSP10_NSP16 5 4 3.42
#> 10 139 298 NSP10 NSP16 NSP10_NSP16 4 5 3.69
#> 11 527 139 NSP14 NSP10 NSP14_NSP10 1 1 4.29
#> 12 527 139 NSP14 NSP10 NSP14_NSP10 2 2 4.28
#> 13 527 139 NSP14 NSP10 NSP14_NSP10 5 3 4.27
#> 14 527 139 NSP14 NSP10 NSP14_NSP10 4 4 4.36
#> 15 527 139 NSP14 NSP10 NSP14_NSP10 3 5 4.37The function pae.boxplot() will plot the lowest average PAE values
from the eight clusters. Each dot represents an AFM predicted structural
model, which typically predicts five models.
pae.boxplot(pae_interface = YOUR_INTERFACE)
The function afm.pisa.heatmap() will plot for all complex, the minimum
average PAE from the eight clusters, and the PDBePISA calculated
solvation free energy (𝚫G), interface area and surface area for all AFM
predicted structural models (typically five per complex).
afm.pisa.heatmap(import_afm = YOUR_AFM_RESULTS, import_pisa = YOUR_PISA_INTERFACE)
#> pae mean of stats::kmeans-clustering: 1 of 15
#> pae mean of stats::kmeans-clustering: 10 of 15
#> Done.
If you provide your own reference set, follow the instructiosn of the binaryPPIclassifier package.
You can also predict the interaction probability of your AFM predicted structural complexes using the maSVM models provided herein, which were trained on 51 known interactions and 67 random protein pairs not known to interact (hsPRS-AF and hsRRS-AF).
data("AFM_maSVM_models")YOUR_TEST_SET <- YOUR_PISA_INTERFACE %>%
left_join(YOUR_INTERFACE %>%
dplyr::select(A_protein, B_protein, complex, model, rank, pae),
by = c("complex" = "complex", "rank" = "rank", "model" = "model")) %>%
dplyr::mutate(deltaG = -deltaG, #invert deltaG
pae = 40-pae) %>% #invert PAE
# adjust columns to meet the binary-PPI-classifier input requirements
dplyr::mutate(Donor = paste(A_protein, rank, sep = "_"),
Donor_tag = "NA",
Donor_protein = A_protein,
Acceptor = paste(B_protein, rank, sep = "_"),
Acceptor_tag = "NA",
Acceptor_protein = B_protein,
complex = "Covid",
reference = "NA",
interaction = paste0(A_protein, " + ", B_protein),
sample = paste0(A_protein, "+", B_protein, "_"),
orientation = paste0(Donor_tag, rank, "+", Acceptor_tag, rank)) %>%
pivot_longer(cols = c(interfaceArea, deltaG, pae),
names_to = "data", values_to = "score")
YOUR_TEST_SET %>% dplyr::select(complex, interaction, orientation, data, score)
#> # A tibble: 45 × 5
#> complex interaction orientation data score
#> <chr> <chr> <chr> <chr> <dbl>
#> 1 Covid E + M NA1+NA1 interfaceArea 1319.
#> 2 Covid E + M NA1+NA1 deltaG 27.2
#> 3 Covid E + M NA1+NA1 pae 24.7
#> 4 Covid E + M NA2+NA2 interfaceArea 967.
#> 5 Covid E + M NA2+NA2 deltaG 20.1
#> 6 Covid E + M NA2+NA2 pae 23.7
#> 7 Covid E + M NA3+NA3 interfaceArea 986.
#> 8 Covid E + M NA3+NA3 deltaG 13.9
#> 9 Covid E + M NA3+NA3 pae 22.2
#> 10 Covid E + M NA4+NA4 interfaceArea 1034.
#> # ℹ 35 more rowsTEST_MAT <- YOUR_TEST_SET %>%
tidyr::unite(complex, interaction, sample, orientation, col = "sample", sep = ";") %>%
tidyr::pivot_wider(names_from = data, values_from = score) %>%
dplyr::filter(across(.cols = c('interfaceArea', 'pae'), ~!is.na(.x))) %>%
tibble::column_to_rownames("sample") %>%
dplyr::select(c('interfaceArea', 'pae')) %>%
base::as.matrix()
TEST_MAT
#> interfaceArea pae
#> Covid;E + M;E+M_;NA1+NA1 1319.1750 24.67004
#> Covid;E + M;E+M_;NA2+NA2 967.4365 23.67244
#> Covid;E + M;E+M_;NA3+NA3 985.7920 22.18190
#> Covid;E + M;E+M_;NA4+NA4 1034.4300 24.76735
#> Covid;E + M;E+M_;NA5+NA5 787.0985 21.04642
#> Covid;NSP10 + NSP16;NSP10+NSP16_;NA1+NA1 954.7190 36.23047
#> Covid;NSP10 + NSP16;NSP10+NSP16_;NA2+NA2 937.9400 36.69947
#> Covid;NSP10 + NSP16;NSP10+NSP16_;NA3+NA3 1128.0500 36.13220
#> Covid;NSP10 + NSP16;NSP10+NSP16_;NA4+NA4 1227.7250 36.58069
#> Covid;NSP10 + NSP16;NSP10+NSP16_;NA5+NA5 1155.9700 36.31416
#> Covid;NSP14 + NSP10;NSP14+NSP10_;NA1+NA1 2269.7900 35.70924
#> Covid;NSP14 + NSP10;NSP14+NSP10_;NA2+NA2 2308.5450 35.71643
#> Covid;NSP14 + NSP10;NSP14+NSP10_;NA3+NA3 2245.2000 35.73414
#> Covid;NSP14 + NSP10;NSP14+NSP10_;NA4+NA4 2293.1800 35.64153
#> Covid;NSP14 + NSP10;NSP14+NSP10_;NA5+NA5 2208.7600 35.62514prediction <- data.frame()
for(i in 1:length(AFM_maSVM_models)) {
tmp <- attr(stats::predict(AFM_maSVM_models[[i]], newdata = TEST_MAT,
decision.values = TRUE, probability = TRUE), "probabilities")
tmp <- tmp %>%
as.data.frame() %>%
rownames_to_column("id") %>%
tidyr::separate(col = "id",
into = c("complex", "interaction", "sample", "orientation"),
sep = ";")
tmp <- cbind(tmp,i)
prediction <- rbind(prediction, tmp)
rm(tmp)
}
YOUR_AFM_PREDICTIONS <- prediction %>%
group_by(interaction, orientation) %>%
dplyr::summarise(probability = mean(`2`))
#> `summarise()` has grouped output by 'interaction'. You can override using the
#> `.groups` argument.
YOUR_AFM_PREDICTIONS
#> # A tibble: 15 × 3
#> # Groups: interaction [3]
#> interaction orientation probability
#> <chr> <chr> <dbl>
#> 1 E + M NA1+NA1 0.792
#> 2 E + M NA2+NA2 0.617
#> 3 E + M NA3+NA3 0.512
#> 4 E + M NA4+NA4 0.715
#> 5 E + M NA5+NA5 0.346
#> 6 NSP10 + NSP16 NA1+NA1 0.985
#> 7 NSP10 + NSP16 NA2+NA2 0.986
#> 8 NSP10 + NSP16 NA3+NA3 0.989
#> 9 NSP10 + NSP16 NA4+NA4 0.992
#> 10 NSP10 + NSP16 NA5+NA5 0.990
#> 11 NSP14 + NSP10 NA1+NA1 0.998
#> 12 NSP14 + NSP10 NA2+NA2 0.998
#> 13 NSP14 + NSP10 NA3+NA3 0.998
#> 14 NSP14 + NSP10 NA4+NA4 0.998
#> 15 NSP14 + NSP10 NA5+NA5 0.997AI-guided pipeline for protein–protein interaction drug discovery identifies a SARS-CoV-2 inhibitor
Trepte P, Secker C, Olivet J, Blavier J, Kostova S, Maseko SB, Minia I, Ramos ES, Cassonnet P, Golusik S, et al (2024) AI-guided pipeline for protein–protein interaction drug discovery identifies a SARS-CoV-2 inhibitor. Mol Syst Biol: 1–30
https://doi.org/10.1038/s44320-024-00019-8
Distributed under the MIT License. See License.md for more
information.
Philipp Trepte - philipp.trepte@imba.oeaw.ac.at - LinkedIn
AFM-PISA-classifier: https://github.com/philipptrepte/AFM-PISA-classifier
sessionInfo()
#> R version 4.3.2 (2023-10-31)
#> Platform: aarch64-apple-darwin20 (64-bit)
#> Running under: macOS Sonoma 14.4
#>
#> Matrix products: default
#> BLAS: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRblas.0.dylib
#> LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.11.0
#>
#> locale:
#> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
#>
#> time zone: Europe/Berlin
#> tzcode source: internal
#>
#> attached base packages:
#> [1] grid stats4 stats graphics grDevices utils datasets
#> [8] methods base
#>
#> other attached packages:
#> [1] knitr_1.45 DT_0.32
#> [3] AFMpisa_1.0.2.0 xml2_1.3.6
#> [5] ComplexHeatmap_2.18.0 Biostrings_2.70.3
#> [7] GenomeInfoDb_1.38.8 XVector_0.42.0
#> [9] IRanges_2.36.0 S4Vectors_0.40.2
#> [11] BiocGenerics_0.48.1 rjson_0.2.21
#> [13] binaryPPIclassifier_1.5.5.8 plotly_4.10.4
#> [15] dplyr_1.1.4 viridis_0.6.5
#> [17] viridisLite_0.4.2 varhandle_2.0.6
#> [19] usethis_2.2.3 tidyr_1.3.1
#> [21] tibble_3.2.1 stringr_1.5.1
#> [23] Rmisc_1.5.1 plyr_1.8.9
#> [25] rlang_1.1.3 randomForest_4.7-1.1
#> [27] purrr_1.0.2 plotROC_2.3.1
#> [29] ggpubr_0.6.0 ggnewscale_0.4.10
#> [31] e1071_1.7-14 DescTools_0.99.54
#> [33] cowplot_1.1.3 caret_6.0-94
#> [35] lattice_0.22-6 ggplot2_3.5.0
#>
#> loaded via a namespace (and not attached):
#> [1] RColorBrewer_1.1-3 rstudioapi_0.15.0 jsonlite_1.8.8
#> [4] shape_1.4.6.1 magrittr_2.0.3 magick_2.8.3
#> [7] farver_2.1.1 rmarkdown_2.26 GlobalOptions_0.1.2
#> [10] fs_1.6.3 zlibbioc_1.48.2 vctrs_0.6.5
#> [13] Cairo_1.6-2 RCurl_1.98-1.14 rstatix_0.7.2
#> [16] htmltools_0.5.7 broom_1.0.5 cellranger_1.1.0
#> [19] pROC_1.18.5 parallelly_1.37.1 htmlwidgets_1.6.4
#> [22] rootSolve_1.8.2.4 lubridate_1.9.3 lifecycle_1.0.4
#> [25] iterators_1.0.14 pkgconfig_2.0.3 Matrix_1.6-5
#> [28] R6_2.5.1 fastmap_1.1.1 clue_0.3-65
#> [31] GenomeInfoDbData_1.2.11 future_1.33.1 digest_0.6.35
#> [34] Exact_3.2 colorspace_2.1-0 labeling_0.4.3
#> [37] fansi_1.0.6 timechange_0.3.0 httr_1.4.7
#> [40] abind_1.4-5 compiler_4.3.2 proxy_0.4-27
#> [43] withr_3.0.0 doParallel_1.0.17 backports_1.4.1
#> [46] carData_3.0-5 highr_0.10 ggsignif_0.6.4
#> [49] MASS_7.3-60.0.1 lava_1.8.0 gld_2.6.6
#> [52] ModelMetrics_1.2.2.2 tools_4.3.2 future.apply_1.11.1
#> [55] nnet_7.3-19 glue_1.7.0 nlme_3.1-164
#> [58] cluster_2.1.6 reshape2_1.4.4 generics_0.1.3
#> [61] recipes_1.0.10 gtable_0.3.4 class_7.3-22
#> [64] data.table_1.15.2 lmom_3.0 car_3.1-2
#> [67] utf8_1.2.4 foreach_1.5.2 pillar_1.9.0
#> [70] circlize_0.4.16 splines_4.3.2 survival_3.5-8
#> [73] tidyselect_1.2.1 gridExtra_2.3 xfun_0.42
#> [76] expm_0.999-9 hardhat_1.3.1 timeDate_4032.109
#> [79] matrixStats_1.2.0 stringi_1.8.3 lazyeval_0.2.2
#> [82] yaml_2.3.8 boot_1.3-30 evaluate_0.23
#> [85] codetools_0.2-19 cli_3.6.2 rpart_4.1.23
#> [88] munsell_0.5.0 Rcpp_1.0.12 readxl_1.4.3
#> [91] globals_0.16.3 png_0.1-8 parallel_4.3.2
#> [94] assertthat_0.2.1 gower_1.0.1 bitops_1.0-7
#> [97] listenv_0.9.1 mvtnorm_1.2-4 ipred_0.9-14
#> [100] scales_1.3.0 prodlim_2023.08.28 crayon_1.5.2
#> [103] GetoptLong_1.0.5