SPCI software calculates fragments contributions to a modeled property by means of interpretation of QSAR models (https://github.com/DrrDom/spci). This package helps to analyze and visualize contributions.

How to install

First install devtools package

instal.packages("devtools")

Then run from the R console

devtools::install_github("DrrDom/rspci")

Use the same command to update the package if necessary.

How to use

Load package and prepare data before visualization

library(rspci)

load text file with contributions (normally this file is located in a dir of modeled property and names default_frag_contributions.txt or auto_frag_contributions.txt, etc)

file_name <- system.file("extdata", "BBB_frag_contributions.txt", package = "rspci")
d <- load_data(file_name)

add full names containing info about fragments occurencies

d <- add_full_names(d)

filter fragments by count to discard rarely occured fragments (optional); there are other filters which may be applied

d <- filter_by_frags_count(d)

reorder data according to specific model and contribution for nice visualization and consistent order of fragments on different plots

d <- reorder_data(d, "consensus", "overall")

add significance levels to data

d <- add_signif(d)

Visualization of overall contributions

keep only overall contributions for visualization

d1 <- filter_by_prop_names(d, "overall")

boxplot

plot_contrib(d1)

barplot

plot_contrib(d1, plot_type = "barplot")

flipped boxplot with significance levels

plot_contrib(d1, flip = FALSE, show_sign_text = "ptext")

barpplot with significance levels

plot_contrib(d1, plot_type = "barplot", show_sign_text = "ptext")

Visualization of physico-chemical contributions

keep only physico-chemical contributions

d2 <- filter_by_prop_names(d, remove_prop_names = "overall")

barplot

plot_contrib(d2, plot_type = "barplot")

not fliped barplot

plot_contrib(d2, plot_type = "barplot", flip = FALSE)

not fliped barplot with siginificance levels

plot_contrib(d2, plot_type = "barplot", flip = FALSE, show_sign_text = "ptext")

Create summary table

These commands return summary table of the consensus model sorted by median overall contributions of the fragments

library(dplyr)  # to use %>% function

d <- add_full_names(d)  # required, as full_name column will be used for grouping
d <- filter_by_frags_count(d)  # optional, to remove rarely occured fragments

df <- d %>%
  filter(Property == "overall", Model == "consensus") %>%
  group_by(full_name) %>%
  summarise(median = median(Contribution)) %>%
  arrange(desc(median))

Clustering fragment contributions to detect specific molecular context

Load data again

file_name <- system.file("extdata", "BBB_frag_contributions.txt", package = "rspci")
d <- load_data(file_name)
d <- add_full_names(d)

select a fragment and desired contributions to cluster

dx <- dplyr::filter(d, FragID == "OH (aliphatic)", Model == "consensus", Property == "overall")

build a Gaussian mixture model which detects possible clusters in the distribution of fragment contributions

m <- clust(dx$Contribution, dx$MolID)

The model can be visualized

plot_mclust(m, "Main title", 0.05)

Parameters (mean, variance, proportion) of the clusters can be retrieved

p <- get_clust_params(m)

Molecule IDs which where supplied to clust function together with fragment contributions can be retrieved as well and further analyzed for possible patterns

ids <- get_mol_ids(m, uncert = 0.2)

It is possible to build many models at once for a selected model and save them to grid image

dy <- dplyr::filter(d, Model == "consensus", Property == "overall")
models <- clust_all(dy, "full_name")
save_mclust_plots("models.png", models)

Citation

1. Polishchuk, P. G.; Kuz'min, V. E.; Artemenko, A. G.; Muratov, E. N., Universal Approach for Structural Interpretation of Qsar/Qspr Models. Mol. Inf. 2013, 32, 843-853 - http://dx.doi.org/10.1002/minf.201300029 - structural interpretation.
2. Polishchuk, P.; Tinkov, O.; Khristova, T.; Ognichenko, L.; Kosinskaya, A.; Varnek, A.; Kuz’min, V., Structural and Physico-Chemical Interpretation (SPCI) of QSAR Models and Its Comparison with Matched Molecular Pair Analysis. J. Chem. Inf. Model. 2016, 56, 1455-1469 - http://dx.doi.org/10.1021/acs.jcim.6b00371 - integrated structural and physicochemical interpretation.
3. Matveieva, M.; Cronin, M. T. D.; Polishchuk, P., Interpretation of QSAR Models: Mining Structural Patterns Taking into Account Molecular Context. Molecular Informatics 2018, 38, 1800084. - https://doi.org/10.1002/minf.201800084 - context-dependent interpretation