2.1 Installation

The R software for running ChemmineR can be downloaded from CRAN (http://cran.at.r-project.org/). The ChemmineR package can be installed from R using the bioLite install command.

 source("http://bioconductor.org/biocLite.R") # Sources the biocLite.R installation script. 
 biocLite("ChemmineR") # Installs the package. 

2.2 Loading the Package and Documentation

 library("ChemmineR") # Loads the package

 library(help="ChemmineR") # Lists all functions and classes 
 vignette("ChemmineR") # Opens this PDF manual from R 

2.3 Five Minute Tutorial

The following code gives an overview of the most important functionalities provided by ChemmineR. Copy and paste of the commands into the R console will demonstrate their utilities.

Create Instances of SDFset class:

 data(sdfsample) 
 sdfset <- sdfsample
 sdfset # Returns summary of SDFset 

## An instance of "SDFset" with 100 molecules

 sdfset[1:4] # Subsetting of object

## An instance of "SDFset" with 4 molecules

 sdfset[[1]] # Returns summarized content of one SDF

## An instance of "SDF"
## 
## <<header>>
##                             Molecule_Name                                    Source 
##                                  "650001"                  "  -OEChem-07071010512D" 
##                                   Comment                               Counts_Line 
##                                        "" " 61 64  0     0  0  0  0  0  0999 V2000" 
## 
## <<atomblock>>
##           C1      C2  C3  C5  C6  C7  C8  C9 C10 C11 C12 C13 C14 C15 C16
## O_1   7.0468  0.0839   0   0   0   0   0   0   0   0   0   0   0   0   0
## O_2  12.2708  1.0492   0   0   0   0   0   0   0   0   0   0   0   0   0
## ...      ...     ... ... ... ... ... ... ... ... ... ... ... ... ... ...
## H_60  1.8411 -1.5985   0   0   0   0   0   0   0   0   0   0   0   0   0
## H_61  2.6597 -1.2843   0   0   0   0   0   0   0   0   0   0   0   0   0
## 
## <<bondblock>>
##      C1  C2  C3  C4  C5  C6  C7
## 1     1  16   2   0   0   0   0
## 2     2  23   1   0   0   0   0
## ... ... ... ... ... ... ... ...
## 63   33  60   1   0   0   0   0
## 64   33  61   1   0   0   0   0
## 
## <<datablock>> (33 data items)
##           PUBCHEM_COMPOUND_CID PUBCHEM_COMPOUND_CANONICALIZED      PUBCHEM_CACTVS_COMPLEXITY 
##                       "650001"                            "1"                          "700" 
##  PUBCHEM_CACTVS_HBOND_ACCEPTOR                                
##                            "7"                          "..."

 view(sdfset[1:4]) # Returns summarized content of many SDFs, not printed here 
 as(sdfset[1:4], "list") # Returns complete content of many SDFs, not printed here 

An SDFset is created during the import of an SD file:

 sdfset <- read.SDFset("http://faculty.ucr.edu/ tgirke/Documents/R_BioCond/Samples/sdfsample.sdf") 

Miscellaneous accessor methods for SDFset container:

 header(sdfset[1:4]) # Not printed here

 header(sdfset[[1]])

##                             Molecule_Name                                    Source 
##                                  "650001"                  "  -OEChem-07071010512D" 
##                                   Comment                               Counts_Line 
##                                        "" " 61 64  0     0  0  0  0  0  0999 V2000"

 atomblock(sdfset[1:4]) # Not printed here 

atomblock(sdfset[[1]])[1:4,] 

##          C1     C2 C3 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16
## O_1  7.0468 0.0839  0  0  0  0  0  0   0   0   0   0   0   0   0
## O_2 12.2708 1.0492  0  0  0  0  0  0   0   0   0   0   0   0   0
## O_3 12.2708 3.1186  0  0  0  0  0  0   0   0   0   0   0   0   0
## O_4  7.9128 2.5839  0  0  0  0  0  0   0   0   0   0   0   0   0

bondblock(sdfset[1:4]) # Not printed here 

 bondblock(sdfset[[1]])[1:4,] 

##   C1 C2 C3 C4 C5 C6 C7
## 1  1 16  2  0  0  0  0
## 2  2 23  1  0  0  0  0
## 3  2 27  1  0  0  0  0
## 4  3 25  1  0  0  0  0

 datablock(sdfset[1:4]) # Not printed here 

 datablock(sdfset[[1]])[1:4] 

##           PUBCHEM_COMPOUND_CID PUBCHEM_COMPOUND_CANONICALIZED      PUBCHEM_CACTVS_COMPLEXITY 
##                       "650001"                            "1"                          "700" 
##  PUBCHEM_CACTVS_HBOND_ACCEPTOR 
##                            "7"

Assigning compound IDs and keeping them unique:

 cid(sdfset)[1:4] # Returns IDs from SDFset object

## [1] "CMP1" "CMP2" "CMP3" "CMP4"

 sdfid(sdfset)[1:4] # Returns IDs from SD file header block

## [1] "650001" "650002" "650003" "650004"

 unique_ids <- makeUnique(sdfid(sdfset))

## [1] "No duplicates detected!"

 cid(sdfset) <- unique_ids 

Converting the data blocks in an SDFset to a matrix:

 blockmatrix <- datablock2ma(datablocklist=datablock(sdfset)) # Converts data block to matrix 
 numchar <- splitNumChar(blockmatrix=blockmatrix) # Splits to numeric and character matrix 
 numchar[[1]][1:2,1:2] # Slice of numeric matrix 

##        PUBCHEM_COMPOUND_CID PUBCHEM_COMPOUND_CANONICALIZED
## 650001               650001                              1
## 650002               650002                              1

 numchar[[2]][1:2,10:11] # Slice of character matrix 

##        PUBCHEM_MOLECULAR_FORMULA PUBCHEM_OPENEYE_CAN_SMILES                                     
## 650001 "C23H28N4O6"              "CC1=CC(=NO1)NC(=O)CCC(=O)N(CC(=O)NC2CCCC2)C3=CC4=C(C=C3)OCCO4"
## 650002 "C18H23N5O3"              "CN1C2=C(C(=O)NC1=O)N(C(=N2)NCCCO)CCCC3=CC=CC=C3"

Compute atom frequency matrix, molecular weight and formula:

 propma <- data.frame(MF=MF(sdfset), MW=MW(sdfset), atomcountMA(sdfset))
 propma[1:4, ] 

##                 MF       MW  C  H N O S F Cl
## 650001  C23H28N4O6 456.4916 23 28 4 6 0 0  0
## 650002  C18H23N5O3 357.4069 18 23 5 3 0 0  0
## 650003 C18H18N4O3S 370.4255 18 18 4 3 1 0  0
## 650004 C21H27N5O5S 461.5346 21 27 5 5 1 0  0

Assign matrix data to data block:

 datablock(sdfset) <- propma 
 datablock(sdfset[1]) 

## $`650001`
##           MF           MW            C            H            N            O            S 
## "C23H28N4O6"   "456.4916"         "23"         "28"          "4"          "6"          "0" 
##            F           Cl 
##          "0"          "0"

String searching in SDFset:

 grepSDFset("650001", sdfset, field="datablock", mode="subset") # Returns summary view of matches. Not printed here.

 grepSDFset("650001", sdfset, field="datablock", mode="index") 

## 1 1 1 1 1 1 1 1 1 
## 1 2 3 4 5 6 7 8 9

Export SDFset to SD file:

 write.SDF(sdfset[1:4], file="sub.sdf", sig=TRUE) 

Plot molecule structure of one or many SDFs:

 plot(sdfset[1:4], print=FALSE) # Plots structures to R graphics device 

 sdf.visualize(sdfset[1:4]) # Compound viewing in web browser 

Structure similarity searching and clustering:

 apset <- sdf2ap(sdfset) # Generate atom pair descriptor database for searching 

 data(apset) # Load sample apset data provided by library. 
 cmp.search(apset, apset[1], type=3, cutoff = 0.3, quiet=TRUE) # Search apset database with single compound. 

##   index    cid    scores
## 1     1 650001 1.0000000
## 2    96 650102 0.3516643
## 3    67 650072 0.3117569
## 4    88 650094 0.3094629
## 5    15 650015 0.3010753

 cmp.cluster(db=apset, cutoff = c(0.65, 0.5), quiet=TRUE)[1:4,] # Binning clustering using variable similarity cutoffs. 

## 
## sorting result...

##       ids CLSZ_0.65 CLID_0.65 CLSZ_0.5 CLID_0.5
## 48 650049         2        48        2       48
## 49 650050         2        48        2       48
## 54 650059         2        54        2       54
## 55 650060         2        54        2       54

4.1 Molecular Structure Data

Classes

• SDFstr: intermediate string class to facilitate SD file import; not important for end user

• SDF: container for single molecule imported from an SD file

• SDFset: container for many SDF objects; most important structure container for end user

• SMI: container for a single SMILES string

• SMIset: container for many SMILES strings

Functions/Methods (mainly for SDFset container, SMIset should be coerced with smiles2sd to SDFset)

• Accessor methods for SDF/SDFset

  • Object slots: cid, header, atomblock, bondblock, datablock (sdfid, datablocktag)

  • Summary of SDFset: view

  • Matrix conversion of data block: datablock2ma, splitNumChar

  • String search in SDFset: grepSDFset

• Coerce one class to another

  • Standard syntax as(..., "...") works in most cases. For details see R help with ?"SDFset-class".

• Utilities

  • Atom frequencies: atomcountMA, atomcount

  • Molecular weight: MW

  • Molecular formula: MF

  • …

• Compound structure depictions

  • R graphics device: plot, plotStruc

  • Online: cmp.visualize

4.2 Structure Descriptor Data

Classes

• AP: container for atom pair descriptors of a single molecule

• APset: container for many AP objects; most important structure descriptor container for end user

• FP: container for fingerprint of a single molecule

• FPset: container for fingerprints of many molecules, most important structure descriptor container for end user

Functions/Methods

• Create AP/APset instances

  • From SDFset: sdf2ap

  • From SD file: cmp.parse

  • Summary of AP/APset: view, db.explain

• Accessor methods for AP/APset

  • Object slots: ap, cid

• Coerce one class to another

  • Standard syntax as(..., "...") works in most cases. For details see R help with ?"APset-class".

• Structure Similarity comparisons and Searching

  • Compute pairwise similarities : cmp.similarity, fpSim

  • Search APset database: cmp.search, fpSim

• AP-based Structure Similarity Clustering

  • Single-linkage binning clustering: cmp.cluster

  • Visualize clustering result with MDS: cluster.visualize

  • Size distribution of clusters: cluster.sizestat
• Folding
  • Fold a descriptor with fold
  • Query the number of times a descriptor has been folded: foldCount
  • Query the number of bits in a descriptor: numBits

5.1 SDF Import

The following gives an overview of the most important import/export functionalities for small molecules provided by ChemmineR. The given example creates an instance of the SDFset class using as sample data set the first 100 compounds from this PubChem SD file (SDF): Compound_00650001_00675000.sdf.gz (ftp://ftp.ncbi.nih.gov/pubchem/Compound/CURRENT-Full/SDF/).

SDFs can be imported with the read.SDFset function:

 sdfset <- read.SDFset("http://faculty.ucr.edu/ tgirke/Documents/R_BioCond/Samples/sdfsample.sdf") 

 data(sdfsample) # Loads the same SDFset provided by the library 
 sdfset <- sdfsample
 valid <- validSDF(sdfset) # Identifies invalid SDFs in SDFset objects 
 sdfset <- sdfset[valid] # Removes invalid SDFs, if there are any 

Import SD file into SDFstr container:

 sdfstr <- read.SDFstr("http://faculty.ucr.edu/ tgirke/Documents/R_BioCond/Samples/sdfsample.sdf") 

Create SDFset from SDFstr class:

 sdfstr <- as(sdfset, "SDFstr") 
 sdfstr

## An instance of "SDFstr" with 100 molecules

 as(sdfstr, "SDFset") 

## An instance of "SDFset" with 100 molecules

5.2 SMILES Import

The read.SMIset function imports one or many molecules from a SMILES file and stores them in a SMIset container. The input file is expected to contain one SMILES string per row with tab-separated compound identifiers at the end of each line. The compound identifiers are optional.

Create sample SMILES file and then import it:

 data(smisample); smiset <- smisample
 write.SMI(smiset[1:4], file="sub.smi") 
 smiset <- read.SMIset("sub.smi")

Inspect content of SMIset:

 data(smisample) # Loads the same SMIset provided by the library 
 smiset <- smisample
 smiset 

## An instance of "SMIset" with 100 molecules

 view(smiset[1:2]) 

## $`650001`
## An instance of "SMI"
## [1] "O=C(NC1CCCC1)CN(c1cc2OCCOc2cc1)C(=O)CCC(=O)Nc1noc(c1)C"
## 
## $`650002`
## An instance of "SMI"
## [1] "O=c1[nH]c(=O)n(c2nc(n(CCCc3ccccc3)c12)NCCCO)C"

Accessor functions:

 cid(smiset[1:4]) 

## [1] "650001" "650002" "650003" "650004"

 smi <- as.character(smiset[1:2])

Create SMIset from named character vector:

 as(smi, "SMIset") 

## An instance of "SMIset" with 2 molecules

6.1 SDF Export

Write objects of classes SDFset/SDFstr/SDF to SD file:

 write.SDF(sdfset[1:4], file="sub.sdf") 

Writing customized SDFset to file containing ChemmineR signature, IDs from SDFset and no data block:

 write.SDF(sdfset[1:4], file="sub.sdf", sig=TRUE, cid=TRUE, db=NULL) 

Example for injecting a custom matrix/data frame into the data block of an SDFset and then writing it to an SD file:

 props <- data.frame(MF=MF(sdfset), MW=MW(sdfset), atomcountMA(sdfset)) 
 datablock(sdfset) <- props
 write.SDF(sdfset[1:4], file="sub.sdf", sig=TRUE, cid=TRUE) 

Indirect export via SDFstr object:

 sdf2str(sdf=sdfset[[1]], sig=TRUE, cid=TRUE) # Uses default components 
 sdf2str(sdf=sdfset[[1]], head=letters[1:4], db=NULL) # Uses custom components for header and data block 

Write SDF, SDFset or SDFstr classes to file:

 write.SDF(sdfset[1:4], file="sub.sdf", sig=TRUE, cid=TRUE, db=NULL)
 write.SDF(sdfstr[1:4], file="sub.sdf") 
 cat(unlist(as(sdfstr[1:4], "list")), file="sub.sdf", sep="") 

6.2 SMILES Export

Write objects of class SMIset to SMILES file with and without compound identifiers:

 data(smisample); smiset <- smisample # Sample data set 

 write.SMI(smiset[1:4], file="sub.smi", cid=TRUE) write.SMI(smiset[1:4], file="sub.smi", cid=FALSE) 

10.1 Loading Data

The functions loadSdf and loadSmiles can be used to load compound data from either a file (both) or an SDFset (loadSdf only). The fct parameter should be a function to extract features from the data. It will be handed an SDFset generated from the data being loaded. This may be done in batches, so there is no guarantee that the given SDFSset will contain the whole dataset. This function should return a data frame with a column for each feature and a row for each compound given. The order of the final data frame should be the same as that of the SDFset. The column names will become the feature names. Each of these features will become a new, indexed, table in the database which can be used later to search for compounds.

The descriptors parameter can be a function which computes descriptors. This function will also be given an SDFset object, which may be done in batches. It should return a data frame with the following two columns: “descriptor” and “descriptor_type”. The “descriptor” column should contain a string representation of the descriptor, and “descriptor_type” is the type of the descriptor. Our convention for atom pair is “ap” and “fp” for finger print. The order should also be maintained.

When the data has been loaded, loadSdf will return the compound id numbers of each compound loaded. These compound id numbers are computed by the database and are not extracted from the compound data itself. They can be used to quickly retrieve compounds later.

New features can also be added using this function. However, all compounds must have all features so if new features are added to a new set of compounds, all existing features must be computable by the fct function given. If new features are detected, all existing compounds will be run through fct in order to compute the new features for them as well.

For example, if dataset X is loaded with features F1 and F2, and then at a later time we load dataset Y with new feature F3, the fct function used to load dataset Y must compute and return features F1, F2, and F3. loadSdf will call fct with both datasets X and Y so that all features are available for all compounds. If any features are missing an error will be raised. If just new features are being added, but no new compounds, use the addNewFeatures function.

In this example, we create a new database called “test.db” and load it with data from an SDFset. We also define fct to compute the molecular weight, “MW”, and the number of rings and aromatic rings. The rings function actually returns a data frame with columns “RINGS” and “AROMATIC”, which will be merged into the data frame being created which will also contain the “MW” column. These will be the names used for these features and must be used when searching with them. Finally, the new compound ids are returned and stored in the “ids” variable.

 data(sdfsample)

 #create and initialize a new SQLite database 
 conn <- initDb("test.db")

## Loading required package: RSQLite
## Loading required package: DBI

## [1] "createing db"

 # load data and compute 3 features: molecular weight, with the MW function, 
 # and counts for RINGS and AROMATIC, as computed by rings, which 
 # returns a data frame itself. 
 ids<-loadSdf(conn,sdfsample, function(sdfset) 
                     data.frame(rings(sdfset,type="count",upper=6, arom=TRUE),propOB(sdfset)) ) 

## adding new features to existing compounds. This could take a while

 #list features in the database:
 print(listFeatures(conn))

##  [1] "aromatic" "cansmi"   "cansmins" "formula"  "hba1"     "hba2"     "hbd"      "inchi"   
##  [9] "logp"     "mr"       "mw"       "nf"       "rings"    "title"    "tpsa"

10.2 Updates

By default the loadSdf / loadSmiles functions will detect duplicate compound entries and only insert one of them. This means it is safe to run these functions on the same data set several times and you won’t end up with duplicates. This allows the functions to be re-run in the event that a previous run on a dataset does not complete. Duplicate compounds are detected by compouting the MD5 checksum on the textual representation of it.

It can also update existing compounds with new versions of the same compound. To enable this, set updateByName to true. It will then consider two compounds with the same name to be the same, even if the definition is different. Then, if the name of a compound exists in the database and it is trying to insert another compound with the same name, it will overwrite the existing compound. It will also drop and re-compute all associated descriptors and features for the new compound (assuming the required functions for descriptor and feature computation are available at the time the update is performed).

10.3 Duplicate Descriptors

It is often the case when loading a large set of compounds that several compounds will produce the same descriptor. ChemmineR detects this case and only stores one copy of the descriptor for every compound it is for. This feature saves some space and some time for processes that need to be applied to every descriptor. It also highlights a new problem. If you have a descriptor in hand and you want to find a single compound to represent it, which compound should be used if the descriptor was produced from multiple compounds? To address this problem, ChemmineR allows you to set priority values for each compound-descriptor mapping. Then, in contexts where a single compound is required, the highest priority compound will be chosen. Highest priority corresponds to the lowest numerical value. So mapping with priority 0 would be used first.

To set these priorities there is the function setPriorities. It takes a function, priorityFn, for computing these priority values. The setPriorities function should be run after loading a complete set of data. It will find each group of compounds which share the same descriptor and call the given function, priorityFn, with the compound_id numbers of the group. This function should then assign priorities to each compound-descriptor pair, however it wishes.

One built in priority function is forestSizePriorities. This simply prefers compounds with fewer disconnected components over compounds with more dissconnected components.

setPriorities(conn,forestSizePriorities)

10.4 Searching

Compounds can be searched for using the findCompounds function. This function takes a connection object, a vector of feature names used in the tests, and finally, a vector of tests that must all pass for a compound to be included in the result set. Each test should be a boolean expression. For example: c("MW <= 400","RINGS \> 3") would return all compounds with a molecular weight of 400 or less and more than 3 rings, assuming these features exist in the database. The syntax for each test is '\<feature name\> \<SQL operator\> \<value\>'. If you know SQL you can go beyond this basic syntax. These tests will simply be concatenated together with “AND” in-between them and tacked on the end of a WHERE clause of an SQL statement. So any SQL that will work in that context is fine. The function will return a list of compound ids, the actual compounds can be fetched with getCompounds. If just the names are needed, the getCompoundNames function can be used. Compounds can also be fetched by name using the findCompoundsByName function.

In this example we search for compounds with molecular weight less than 300.

results = findCompounds(conn,"mw",c("mw < 300"))
message("found ",length(results))

## found 23

If more than one test is given, only compounds which satisfy all tests are found. So if we wanted to further restrict our search to compounds with 2 or more aromatic rings we could do:

results = findCompounds(conn,c("mw","aromatic"),c("mw < 300","aromatic >= 2"))
message("found ",length(results))

## found 9

Remember that any feature used in some test must be listed in the second argument.

String patterns can also be used. So if we wanted to match a substring of the molecular formula, say to find compounds with 21 carbon atoms, we could do:

results = findCompounds(conn,"formula",c("formula like '%C21%'"))
message("found ",length(results))

## found 12

The “like” operator does a pattern match. There are two wildcard operators that can be used with this operator. The “%” will match any stretch of characters while the “?” will match any single character. So the above expression would match a formula like “C21H28N4O6”.

Valid comparison operators are:

• <, <=, > , >=
• =, ==, !=, <>, IS, IS NOT, IN, LIKE

The boolean operators “AND” and “OR” can also be used to create more complex expressions within a single test.

If you just want to fetch every compound in the database you can use the getAllCompoundIds function:

allIds = getAllCompoundIds(conn)
message("found ",length(allIds))

## found 100

10.5 Using Search Results

Once you have a list of compound ids from the findCompounds function, you can either fetch the compound names, or the whole set of compounds as an SDFset.

#get the names of the compounds:
names = getCompoundNames(conn,results)

#if the name order is important set keepOrder=TRUE 
#It will take a little longer though
names = getCompoundNames(conn,results,keepOrder=TRUE) 


# get the whole set of compounds
compounds = getCompounds(conn,results)
#in order:
compounds = getCompounds(conn,results,keepOrder=TRUE)
#write results directly to a file:
compounds = getCompounds(conn,results,filename=file.path(tempdir(),"results.sdf"))

Using the getCompoundFeatures function, you can get a set of feature values as a data frame:

getCompoundFeatures(conn,results[1:5],c("mw","logp","formula"))

##   compound_id       mw   logp       formula
## 1         204 461.5346 1.3127   C21H27N5O5S
## 2         208 438.4780 3.5492  C21H19FN6O2S
## 3         221 340.4592 3.1325    C21H28N2O2
## 4         238 447.9351 5.2940 C21H22ClN3O4S
## 5         243 456.5181 3.0020   C21H24N6O4S

#write results directly to a CSV file (reduces memory usage):
getCompoundFeatures(conn,results[1:5],c("mw","logp","formula"),filename="features.csv")

#maintain input order in output:
print(results[1:5])

## [1] 204 208 221 238 243

getCompoundFeatures(conn,results[1:5],c("mw","logp","formula"),keepOrder=TRUE)

##     compound_id       mw   logp       formula
## 204         204 461.5346 1.3127   C21H27N5O5S
## 208         208 438.4780 3.5492  C21H19FN6O2S
## 221         221 340.4592 3.1325    C21H28N2O2
## 238         238 447.9351 5.2940 C21H22ClN3O4S
## 243         243 456.5181 3.0020   C21H24N6O4S

10.6 Pre-Built Databases

We have pre-built SQLite databases for the Drug Bank and DUD datasets. They can be found in the ChemmineDrugs annotation package. Connections to these databases can be fetched from the functions DrugBank and DUD to get the corresponding database. Any of the above functions can then be used to query the database.

The DUD dataset was downloaded from here. A description can be found here.

The Drug Bank data set is version 4.1. It can be downloaded here

The following features are included:

• aromatic: Number of aromatic rings
• cansmi: Canonical SMILES sting
• cansmins:
• formula: Molecular formula
• hba1:
• hba2:
• hbd:
• inchi: INCHI string
• logp:
• mr:
• mw: Molecular weight
• ncharges:
• nf:
• r2nh:
• r3n:
• rcch:
• rcho:
• rcn:
• rcooh:
• rcoor:
• rcor:
• rings:
• rnh2:
• roh:
• ropo3:
• ror:
• title:
• tpsa:

The DUD database additionally includes:

• target_name: Name of the target
• type: either “active” or “decoy”

16.1 R Graphics Device

A new plotting function for compound structures has been added to the package recently. This function uses the native R graphics device for generating compound depictions. At this point this function is still in an experimental developmental stage but should become stable soon.

If you have ChemmineOB available you can use the regenCoords option to have OpenBabel regenerate the coordinates for the compound. This can sometimes produce better looking plots.

Plot compound Structures with R’s graphics device:

 data(sdfsample)
 sdfset <- sdfsample
 plot(sdfset[1:4], regenCoords=TRUE,print=FALSE) # 'print=TRUE' returns SDF summaries

Customized plots:

 plot(sdfset[1:4], griddim=c(2,2), print_cid=letters[1:4], print=FALSE, 
        noHbonds=FALSE) 

In the following plot, the atom block position numbers in the SDF are printed next to the atom symbols (atomnum = TRUE). For more details, consult help documentation with ?plotStruc or ?plot.

 plot(sdfset["CMP1"], atomnum = TRUE, noHbonds=F , no_print_atoms = "",
        atomcex=0.8, sub=paste("MW:", MW(sdfsample["CMP1"])), print=FALSE) 

Substructure highlighting by atom numbers:

 plot(sdfset[1], print=FALSE, colbonds=c(22,26,25,3,28,27,2,23,21,18,8,19,20,24)) 

16.2 Online with ChemMine Tools

Alternatively, one can visualize compound structures with a standard web browser using the online ChemMine Tools service.

Plot structures using web service ChemMine Tools:

 sdf.visualize(sdfset[1:4]) 

17.1 Maximum Common Substructure (MCS) Searching

The ChemmineR add-on package fmcsR provides support for identifying maximum common substructures (MCSs) and flexible MCSs among compounds. The algorithm can be used for pairwise compound comparisons, structure similarity searching and clustering. The manual describing this functionality is available here and the associated publication is Wang et al. (2013). The following gives a short preview of some functionalities provided by the fmcsR package.

 library(fmcsR)
 data(fmcstest) # Loads test sdfset object 
 test <- fmcs(fmcstest[1], fmcstest[2], au=2, bu=1) # Searches for MCS with mismatches 
 plotMCS(test) # Plots both query compounds with MCS in color 

17.2 AP/APset Classes for Storing Atom Pair Descriptors

The function sdf2ap computes atom pair descriptors for one or many compounds (Carhart, Smith, and Venkataraghavan 1985; Chen and Reynolds 2002). It returns a searchable atom pair database stored in a container of class APset, which can be used for structural similarity searching and clustering. As similarity measure, the Tanimoto coefficient or related coefficients can be used. An APset object consists of one or many AP entries each storing the atom pairs of a single compound. Note: the deprecated cmp.parse function is still available which also generates atom pair descriptor databases, but directly from an SD file. Since the latter function is less flexible it may be discontinued in the future.

Generate atom pair descriptor database for searching:

 ap <- sdf2ap(sdfset[[1]]) # For single compound
 ap 

## An instance of "AP"
## <<atom pairs>>
## 52614450304 52615497856 52615514112 52616547456 52616554624 ... length: 528

 apset <- sdf2ap(sdfset)
 # For many compounds. 

view(apset[1:4]) 

## $`650001`
## An instance of "AP"
## <<atom pairs>>
## 53688190976 53688190977 53688190978 53688190979 53688190980 ... length: 528 
## 
## $`650002`
## An instance of "AP"
## <<atom pairs>>
## 53688190976 53688190977 53688190978 53688190979 53689239552 ... length: 325 
## 
## $`650003`
## An instance of "AP"
## <<atom pairs>>
## 52615496704 53688190976 53688190977 53689239552 53697627136 ... length: 325 
## 
## $`650004`
## An instance of "AP"
## <<atom pairs>>
## 52617593856 52618642432 52619691008 52619691009 52628079616 ... length: 496

Return main components of APset objects:

 cid(apset[1:4]) # Compound IDs 
 ap(apset[1:4]) # Atom pair
 descriptors 
 db.explain(apset[1]) # Return atom pairs in human readable format 

Coerce APset to other objects:

 apset2descdb(apset) # Returns old list-style AP database 
 tmp <- as(apset, "list") # Returns list 
 as(tmp, "APset") # Converts list back to APset 

17.3 Large SDF and Atom Pair Databases

When working with large data sets it is often desirable to save the SDFset and APset containers as binary R objects to files for later use. This way they can be loaded very quickly into a new R session without recreating them every time from scratch.

Save and load of SDFset and APset containers:

 save(sdfset, file = "sdfset.rda", compress = TRUE) 
 load("sdfset.rda") save(apset, file = "apset.rda", compress = TRUE)
 load("apset.rda") 

17.4 Pairwise Compound Comparisons with Atom Pairs

The cmp.similarity function computes the atom pair similarity between two compounds using the Tanimoto coefficient as similarity measure. The coefficient is defined as c/(a+b+c), which is the proportion of the atom pairs shared among two compounds divided by their union. The variable c is the number of atom pairs common in both compounds, while a and b are the numbers of their unique atom pairs.

 cmp.similarity(apset[1],
 apset[2])

## [1] 0.2637037

 cmp.similarity(apset[1], apset[1]) 

## [1] 1

17.5 Similarity Searching with Atom Pairs

The cmp.search function searches an atom pair database for compounds that are similar to a query compound. The following example returns a data frame where the rows are sorted by the Tanimoto similarity score (best to worst). The first column contains the indices of the matching compounds in the database. The argument cutoff can be a similarity cutoff, meaning only compounds with a similarity value larger than this cutoff will be returned; or it can be an integer value restricting how many compounds will be returned. When supplying a cutoff of 0, the function will return the similarity values for every compound in the database.

 cmp.search(apset,
 apset["650065"], type=3, cutoff = 0.3, quiet=TRUE) 

##   index    cid    scores
## 1    61 650066 1.0000000
## 2    60 650065 1.0000000
## 3    67 650072 0.3389831
## 4    11 650011 0.3190608
## 5    15 650015 0.3184524
## 6    86 650092 0.3154270
## 7    64 650069 0.3010279

Alternatively, the function can return the matches in form of an index or a named vector if the type argument is set to 1 or 2, respectively.

 cmp.search(apset, apset["650065"], type=1, cutoff = 0.3, quiet=TRUE)

## [1] 61 60 67 11 15 86 64

 cmp.search(apset, apset["650065"], type=2, cutoff = 0.3, quiet=TRUE) 

##    650066    650065    650072    650011    650015    650092    650069 
## 1.0000000 1.0000000 0.3389831 0.3190608 0.3184524 0.3154270 0.3010279

17.6 FP/FPset Classes for Storing Fingerprints

The FPset class stores fingerprints of small molecules in a matrix-like representation where every molecule is encoded as a fingerprint of the same type and length. The FPset container acts as a searchable database that contains the fingerprints of many molecules. The FP container holds only one fingerprint. Several constructor and coerce methods are provided to populate FP/FPset containers with fingerprints, while supporting any type and length of fingerprints. For instance, the function desc2fp generates fingerprints from an atom pair database stored in an APset, and as(matrix, "FPset") and as(character, "FPset") construct an FPset database from objects where the fingerprints are represented as matrix or character objects, respectively.

Show slots of FPset class:

 showClass("FPset") 

## Class "FPset" [package "ChemmineR"]
## 
## Slots:
##                                     
## Name:       fpma      type foldCount
## Class:    matrix character   numeric

Instance of FPset class:

 data(apset) 
 fpset <- desc2fp(apset)
 view(fpset[1:2]) 

## $`650001`
## An instance of "FP" of type "unknown-1938"
## <<fingerprint>>
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ... length: 1024 
## 
## $`650002`
## An instance of "FP" of type "unknown-1859"
## <<fingerprint>>
## 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 0 0 0 1 1 ... length: 1024

FPset class usage:

 fpset[1:4] # behaves like a list 

## An instance of a 1024 bit "FPset" of type "apfp" with 4 molecules

 fpset[[1]] # returns FP object 

## An instance of "FP" of type "unknown-3570"
## <<fingerprint>>
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ... length: 1024

 length(fpset) # number of compounds ENDCOMMENT

## [1] 100

 cid(fpset) # returns compound ids 

##   [1] "650001" "650002" "650003" "650004" "650005" "650006" "650007" "650008" "650009" "650010"
##  [11] "650011" "650012" "650013" "650014" "650015" "650016" "650017" "650019" "650020" "650021"
##  [21] "650022" "650023" "650024" "650025" "650026" "650027" "650028" "650029" "650030" "650031"
##  [31] "650032" "650033" "650034" "650035" "650036" "650037" "650038" "650039" "650040" "650041"
##  [41] "650042" "650043" "650044" "650045" "650046" "650047" "650048" "650049" "650050" "650052"
##  [51] "650054" "650056" "650058" "650059" "650060" "650061" "650062" "650063" "650064" "650065"
##  [61] "650066" "650067" "650068" "650069" "650070" "650071" "650072" "650073" "650074" "650075"
##  [71] "650076" "650077" "650078" "650079" "650080" "650081" "650082" "650083" "650085" "650086"
##  [81] "650087" "650088" "650089" "650090" "650091" "650092" "650093" "650094" "650095" "650096"
##  [91] "650097" "650098" "650099" "650100" "650101" "650102" "650103" "650104" "650105" "650106"

 fpset[10] <- 0 # replacement of 10th fingerprint to all zeros 
 cid(fpset) <- 1:length(fpset) # replaces compound ids 
 c(fpset[1:4], fpset[11:14]) # concatenation of several FPset objects 

## An instance of a 1024 bit "FPset" of type "apfp" with 8 molecules

Construct FPset class form matrix:

 fpma <- as.matrix(fpset) # coerces FPset to matrix 
 as(fpma, "FPset") 

## An instance of a 1024 bit "FPset" of type "unknown-4880" with 100 molecules

Construct FPset class form character vector:

 fpchar <- as.character(fpset) # coerces FPset to character strings 
 as(fpchar, "FPset") # construction of FPset class from character vector

## An instance of a 1024 bit "FPset" of type "apfp" with 100 molecules

Compound similarity searching with FPset:

 fpSim(fpset[1], fpset, method="Tanimoto", cutoff=0.4, top=4) 

##         1        96        67        15 
## 1.0000000 0.4719101 0.4288499 0.4275229

Folding fingerprints:

 fold(fpset) # fold each FP once

## An instance of a 512 bit "FPset" of type "apfp" with 100 molecules

 fold(fpset, count=2) #fold each FP twice

## An instance of a 256 bit "FPset" of type "apfp" with 100 molecules

 fold(fpset, bits=128) #fold each FP down to 128 bits

## An instance of a 128 bit "FPset" of type "apfp" with 100 molecules

 fold(fpset[[1]])  # fold an individual FP

## An instance of "FP" of type "unknown-274"
## <<fingerprint>>
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ... length: 512

 fptype(fpset) # get type of FPs

## [1] "apfp"

 numBits(fpset) # get the number of bits of each FP

## [1] 1024

 foldCount(fold(fpset)) # the number of times an FP or FPset has been folded

## [1] 1

17.7 Atom Pair Fingerprints

Atom pairs can be converted into binary atom pair fingerprints of fixed length. Computations on this compact data structure are more time and memory efficient than on their relatively complex atom pair counterparts. The function desc2fp generates fingerprints from descriptor vectors of variable length such as atom pairs stored in APset or list containers. The obtained fingerprints can be used for structure similarity comparisons, searching and clustering.

Create atom pair sample data set:

 data(sdfsample)
 sdfset <- sdfsample[1:10] 
 apset <- sdf2ap(sdfset) 

Compute atom pair fingerprint database using internal atom pair selection containing the 4096 most common atom pairs identified in DrugBank’s compound collection. For details see ?apfp. The following example uses from this set the 1024 most frequent atom pairs:

 fpset <- desc2fp(apset, descnames=1024, type="FPset") 

Alternatively, one can provide any custom atom pair selection. Here, the 1024 most common ones in apset:

 fpset1024 <- names(rev(sort(table(unlist(as(apset, "list")))))[1:1024]) 
 fpset <- desc2fp(apset, descnames=fpset1024, type="FPset") 

A more compact way of storing fingerprints is as character values:

 fpchar <- desc2fp(x=apset,
 descnames=1024, type="character") fpchar <- as.character(fpset) 

Converting a fingerprint database to a matrix and vice versa:

 fpma <- as.matrix(fpset)
 fpset <- as(fpma, "FPset") 

Similarity searching and returning Tanimoto similarity coefficients:

 fpSim(fpset[1], fpset, method="Tanimoto") 

Under method one can choose from several predefined similarity measures including Tanimoto (default), Euclidean, Tversky or Dice. Alternatively, one can pass on custom similarity functions.

 fpSim(fpset[1], fpset, method="Tversky", cutoff=0.4, top=4, alpha=0.5, beta=1) 

Example for using a custom similarity function:

 myfct <- function(a, b, c, d) c/(a+b+c+d)
 fpSim(fpset[1], fpset, method=myfct) 

Clustering example:

 simMAap <- sapply(cid(apfpset), function(x) fpSim(x=apfpset[x], apfpset, sorted=FALSE)) 
 hc <- hclust(as.dist(1-simMAap), method="single")
 plot(as.dendrogram(hc), edgePar=list(col=4, lwd=2), horiz=TRUE) 

17.8 Fingerprint E-values

The fpSim function can also return Z-scores, E-values, and p-values if given a set of score distribution parameters. These parameters can be computed over an fpSet with the genParameters function.

    params <- genParameters(fpset)

This function will compute all pairwise distances between the given fingerprints and then fit a Beta distribution to the resulting Tanimoto scores, conditioned on the number of set bits in each fingerprint. For large data sets where you would not want to compute all pairwise distances, you can set what fraction to sample with the sampleFraction argument. This step only needs to be done once for each database of fpSet objects. Alternatively, if you have a large database of fingerprints, or you believe that the parameters computed on a single database are more generally applicable, you can use the resulting parameters for other databases as well.

Once you have a set of parameters, you can pass them to fpSim with the parameters argument.

    fpSim(fpset[[1]], fpset, top=10, parameters=params) 

##    similarity    zscore    evalue    pvalue
## 1   1.0000000 6.2418215  0.000000 0.0000000
## 96  0.4719101 1.6075792  6.748413 0.9988273
## 67  0.4288499 1.2297052 12.012285 0.9999939
## 15  0.4275229 1.2180604 12.211967 0.9999950
## 88  0.4247423 1.1936587 12.638193 0.9999968
## 64  0.4187380 1.1409688 13.594938 0.9999988
## 4   0.4166667 1.1227914 13.936692 0.9999991
## 86  0.3978686 0.9578290 17.319191 1.0000000
## 77  0.3970588 0.9507232 17.476453 1.0000000
## 69  0.3940000 0.9238806 18.079243 1.0000000

This will then return a data frame with the similarity, Z-score, E-value, and p-value. You can change which value will be used as a cutoff and to sort by by setting the argument scoreType to one of these scores. In this way you could set an E-value cutoff of 0.04 for example.

    fpSim(fpset[[1]], fpset, cutoff=0.04, scoreType="evalue", parameters=params)    

##   similarity   zscore evalue pvalue
## 1          1 6.241822      0      0

17.9 Pairwise Compound Comparisons with PubChem Fingerprints

The fpSim function computes the similarity coefficients (e.g. Tanimoto) for pairwise comparisons of binary fingerprints. For this data type, c is the number of “on-bits” common in both compounds, and a and b are the numbers of their unique “on-bits”. Currently, the PubChem fingerprints need to be provided (here PubChem’s SD files) and cannot be computed from scratch in ChemmineR. The PubChem fingerprint specifications can be loaded with data(pubchemFPencoding).

Convert base 64 encoded PubChem fingerprints to character vector, matrix or FPset object:

 cid(sdfset) <- sdfid(sdfset)
 fpset <- fp2bit(sdfset, type=1) 
 fpset <- fp2bit(sdfset, type=2) 
 fpset <- fp2bit(sdfset, type=3) 
 fpset 

## An instance of a 881 bit "FPset" of type "pubchem" with 100 molecules

Pairwise compound structure comparisons:

 fpSim(fpset[1], fpset[2]) 

##    650002 
## 0.5364807

17.10 Similarity Searching with PubChem Fingerprints

Similarly, the fpSim function provides search functionality for PubChem fingerprints:

 fpSim(fpset["650065"], fpset, method="Tanimoto", cutoff=0.6, top=6) 

##    650065    650066    650035    650019    650012    650046 
## 1.0000000 0.9944751 0.7435897 0.7432432 0.7230047 0.7142857

17.11 Visualize Similarity Search Results

The cmp.search function allows to visualize the chemical structures for the search results. Similar but more flexible chemical structure rendering functions are plot and sdf.visualize described above. By setting the visualize argument in cmp.search to TRUE, the matching compounds and their scores can be visualized with a standard web browser. Depending on the visualize.browse argument, an URL will be printed or a webpage will be opened showing the structures of the matching compounds.

View similarity search results in R’s graphics device:

 cid(sdfset) <-
 cid(apset) # Assure compound name consistency among objects. 

 plot(sdfset[names(cmp.search(apset, apset["650065"], type=2, cutoff=4, quiet=TRUE))], print=FALSE) 

View results online with Chemmine Tools:

 similarities <- cmp.search(apset, apset[1], type=3, cutoff = 10)
 sdf.visualize(sdfset[similarities[,1]]) 

18.1 Clustering Identical or Very Similar Compounds

Often it is of interest to identify very similar or identical compounds in a compound set. The cmp.duplicated function can be used to quickly identify very similar compounds in atom pair sets, which will be frequently, but not necessarily, identical compounds.

Identify compounds with identical AP sets:

 cmp.duplicated(apset, type=1)[1:4] # Returns AP duplicates as logical vector 

## [1] FALSE FALSE FALSE FALSE

 cmp.duplicated(apset, type=2)[1:4,] # Returns AP duplicates as data frame 

##      ids CLSZ_100 CLID_100
## 1 650082        1        1
## 2 650059        2        2
## 3 650060        2        2
## 4 650010        1        3

Plot the structure of two pairs of duplicates:

 plot(sdfset[c("650059","650060", "650065", "650066")], print=FALSE) 

Remove AP duplicates from SDFset and APset objects:

 apdups <- cmp.duplicated(apset, type=1)
 sdfset[which(!apdups)]; apset[which(!apdups)] 

## An instance of "SDFset" with 96 molecules

## An instance of "APset" with 96 molecules

Alternatively, one can identify duplicates via other descriptor types if they are provided in the data block of an imported SD file. For instance, one can use here fingerprints, InChI, SMILES or other molecular representations. The following examples show how to enumerate by identical InChI strings, SMILES strings and molecular formula, respectively.

 count <- table(datablocktag(sdfset,
 tag="PUBCHEM_NIST_INCHI"))
 count <- table(datablocktag(sdfset, tag="PUBCHEM_OPENEYE_CAN_SMILES")) 
 count <- table(datablocktag(sdfset, tag="PUBCHEM_MOLECULAR_FORMULA")) 
 count[1:4] 

## 
##  C10H9FN2O2S   C11H12N4OS    C11H13NO4 C12H11ClN2OS 
##            1            1            1            1

18.2 Binning Clustering

Compound libraries can be clustered into discrete similarity groups with the binning clustering function cmp.cluster. The function accepts as input an atom pair (APset) or a fingerprint (FPset) descriptor database as well as a similarity threshold. The binning clustering result is returned in form of a data frame. Single linkage is used for cluster joining. The function calculates the required compound-to-compound distance information on the fly, while a memory-intensive distance matrix is only created upon user request via the save.distances argument (see below).

Because an optimum similarity threshold is often not known, the cmp.cluster function can calculate cluster results for multiple cutoffs in one step with almost the same speed as for a single cutoff. This can be achieved by providing several cutoffs under the cutoff argument. The clustering results for the different cutoffs will be stored in one data frame.

One may force the cmp.cluster function to calculate and store the distance matrix by supplying a file name to the save.distances argument. The generated distance matrix can be loaded and passed on to many other clustering methods available in R, such as the hierarchical clustering function hclust (see below).

If a distance matrix is available, it may also be supplied to cmp.cluster via the use.distances argument. This is useful when one has a pre-computed distance matrix either from a previous call to cmp.cluster or from other distance calculation subroutines.

Single-linkage binning clustering with one or multiple cutoffs:

 clusters <- cmp.cluster(db=apset, cutoff = c(0.7, 0.8, 0.9), quiet = TRUE)

## 
## sorting result...

 clusters[1:12,] 

##       ids CLSZ_0.7 CLID_0.7 CLSZ_0.8 CLID_0.8 CLSZ_0.9 CLID_0.9
## 48 650049        2       48        2       48        2       48
## 49 650050        2       48        2       48        2       48
## 54 650059        2       54        2       54        2       54
## 55 650060        2       54        2       54        2       54
## 56 650061        2       56        2       56        2       56
## 57 650062        2       56        2       56        2       56
## 58 650063        2       58        2       58        2       58
## 59 650064        2       58        2       58        2       58
## 60 650065        2       60        2       60        2       60
## 61 650066        2       60        2       60        2       60
## 1  650001        1        1        1        1        1        1
## 2  650002        1        2        1        2        1        2

Clustering of FPset objects with multiple cutoffs. This method allows to call various similarity methods provided by the fpSim function. For details consult ?fpSim.

 fpset <- desc2fp(apset)
 clusters2 <- cmp.cluster(fpset, cutoff=c(0.5, 0.7, 0.9), method="Tanimoto", quiet=TRUE)

## 
## sorting result...

 clusters2[1:12,] 

##       ids CLSZ_0.5 CLID_0.5 CLSZ_0.7 CLID_0.7 CLSZ_0.9 CLID_0.9
## 69 650074       14       11        2       69        1       69
## 79 650085       14       11        2       69        1       79
## 11 650011       14       11        1       11        1       11
## 15 650015       14       11        1       15        1       15
## 45 650046       14       11        1       45        1       45
## 47 650048       14       11        1       47        1       47
## 51 650054       14       11        1       51        1       51
## 53 650058       14       11        1       53        1       53
## 64 650069       14       11        1       64        1       64
## 65 650070       14       11        1       65        1       65
## 67 650072       14       11        1       67        1       67
## 86 650092       14       11        1       86        1       86

Sames as above, but using Tversky similarity measure:

 clusters3 <- cmp.cluster(fpset, cutoff=c(0.5, 0.7, 0.9), 
                                  method="Tversky", alpha=0.3, beta=0.7, quiet=TRUE) 

## 
## sorting result...

Return cluster size distributions for each cutoff:

 cluster.sizestat(clusters, cluster.result=1)

##   cluster size count
## 1            1    90
## 2            2     5

 cluster.sizestat(clusters, cluster.result=2)

##   cluster size count
## 1            1    90
## 2            2     5

 cluster.sizestat(clusters, cluster.result=3) 

##   cluster size count
## 1            1    90
## 2            2     5

Enforce calculation of distance matrix:

 clusters <- cmp.cluster(db=apset, cutoff = c(0.65, 0.5, 0.3),
 save.distances="distmat.rda") # Saves distance matrix to file "distmat.rda" in current working directory. 
 load("distmat.rda") # Loads distance matrix. 

18.3 Jarvis-Patrick Clustering

The Jarvis-Patrick clustering algorithm is widely used in cheminformatics (Jarvis and Patrick 1973). It requires a nearest neighbor table, which consists of j nearest neighbors for each item (e.g. compound). The nearest neighbor table is then used to join items into clusters when they meet the following requirements: (a) they are contained in each other’s neighbor list and (b) they share at least k nearest neighbors. The values for j and k are user-defined parameters. The jarvisPatrick function implemented in ChemmineR takes a nearest neighbor table generated by nearestNeighbors, which works for APset and FPset objects. This function takes either the standard Jarvis-Patrick j parameter (as the numNbrs parameter), or else a cutoff value, which is an extension to the basic algorithm that we have added. Given a cutoff value, the nearest neighbor table returned contains every neighbor with a similarity greater than the cutoff value, for each item. This allows one to generate tighter clusters and to minimize certain limitations of this method, such as false joins of completely unrelated items when operating on small data sets. The trimNeighbors function can also be used to take an existing nearest neighbor table and remove all neighbors whose similarity value is below a given cutoff value. This allows one to compute a very relaxed nearest neighbor table initially, and then quickly try different refinements later.

In case an existing nearest neighbor matrix needs to be used, the fromNNMatrix function can be used to transform it into the list structure that jarvisPatrick requires. The input matrix must have a row for each compound, and each row should be the index values of the neighbors of compound represented by that row. The names of each compound can also be given through the names argument. If not given, it will attempt to use the rownames of the given matrix.

The jarvisPatrick function also allows one to relax some of the requirements of the algorithm through the mode parameter. When set to “a1a2b”, then all requirements are used. If set to “a1b”, then (a) is relaxed to a unidirectional requirement. Lastly, if mode is set to “b”, then only requirement (b) is used, which means that all pairs of items will be checked to see if (b) is satisfied between them. The size of the clusters generated by the different methods increases in this order: “a1a2b” < “a1b” < “b”. The run time of method “a1a2b” follows a close to linear relationship, while it is nearly quadratic for the much more exhaustive method “b”. Only methods “a1a2b” and “a1b” are suitable for clustering very large data sets (e.g. >50,000 items) in a reasonable amount of time.

An additional extension to the algorithm is the ability to set the linkage mode. The linkage parameter can be one of “single”, “average”, or “complete”, for single linkage, average linkage and complete linkage merge requirements, respectively. In the context of Jarvis-Patrick, average linkage means that at least half of the pairs between the clusters under consideration must meet requirement (b). Similarly, for complete linkage, all pairs must requirement (b). Single linkage is the normal case for Jarvis-Patrick and just means that at least one pair must meet requirement (b).

The output is a cluster vector with the item labels in the name slot and the cluster IDs in the data slot. There is a utility function called byCluster, which takes out cluster vector output by jarvisPatrick and transforms it into a list of vectors. Each slot of the list is named with a cluster id and the vector contains the cluster members. By default the function excludes singletons from the output, but they can be included by setting excludeSingletons=FALSE`.

Load/create sample APset and FPset:

 data(apset) 
 fpset <- desc2fp(apset) 

Standard Jarvis-Patrick clustering on APset and FPset objects:

 jarvisPatrick(nearestNeighbors(apset,numNbrs=6), k=5, mode="a1a2b")

## 650001 650002 650003 650004 650005 650006 650007 650008 650009 650010 650011 650012 650013 650014 
##      1      2      3      4      5      6      7      8      9     10     11     12     13     14 
## 650015 650016 650017 650019 650020 650021 650022 650023 650024 650025 650026 650027 650028 650029 
##     11     15     16     17     18     19     20     21     22     23     24     25     26     27 
## 650030 650031 650032 650033 650034 650035 650036 650037 650038 650039 650040 650041 650042 650043 
##     28     29     30     31     32     33     34     35     36     37     38     39     40     41 
## 650044 650045 650046 650047 650048 650049 650050 650052 650054 650056 650058 650059 650060 650061 
##     42     43     44     45     46     47     48     49     50     51     52     53     54     55 
## 650062 650063 650064 650065 650066 650067 650068 650069 650070 650071 650072 650073 650074 650075 
##     56     57     58     59     60     61     62     63     64     65     66     67     68     69 
## 650076 650077 650078 650079 650080 650081 650082 650083 650085 650086 650087 650088 650089 650090 
##     70     71     72     73     74     75     76     77     78     79     80     81     82     83 
## 650091 650092 650093 650094 650095 650096 650097 650098 650099 650100 650101 650102 650103 650104 
##     84     85     86     87     88     89     90     91     92     93     94     95     96     97 
## 650105 650106 
##     98     99

 #Using "APset" 

 jarvisPatrick(nearestNeighbors(fpset,numNbrs=6), k=5, mode="a1a2b")

## 650001 650002 650003 650004 650005 650006 650007 650008 650009 650010 650011 650012 650013 650014 
##      1      2      3      4      5      6      7      8      9     10     11     12     13     14 
## 650015 650016 650017 650019 650020 650021 650022 650023 650024 650025 650026 650027 650028 650029 
##     11     15     16     17     18     19     20     21     22     23     24     25     26     27 
## 650030 650031 650032 650033 650034 650035 650036 650037 650038 650039 650040 650041 650042 650043 
##     28     29     30     31     32     33     34     35     36     37     38     39     40     41 
## 650044 650045 650046 650047 650048 650049 650050 650052 650054 650056 650058 650059 650060 650061 
##     42     43     44     45     46     47     48     49     50     51     52     53     54     55 
## 650062 650063 650064 650065 650066 650067 650068 650069 650070 650071 650072 650073 650074 650075 
##     56     57     58     59     60     61     62     63     64     65     66     67     68     69 
## 650076 650077 650078 650079 650080 650081 650082 650083 650085 650086 650087 650088 650089 650090 
##     70     71     72     73     74     75     76     77     78     79     80     81     82     83 
## 650091 650092 650093 650094 650095 650096 650097 650098 650099 650100 650101 650102 650103 650104 
##     84     85     86     87     88     89     90     91     92     93     94      1     95     96 
## 650105 650106 
##     97     98

 #Using "FPset" 

The following example runs Jarvis-Patrick clustering with a minimum similarity cutoff value (here Tanimoto coefficient). In addition, it uses the much more exhaustive "b" method that generates larger cluster sizes, but significantly increased the run time. For more details, consult the corresponding help file with ?jarvisPatrick.

 cl<-jarvisPatrick(nearestNeighbors(fpset,cutoff=0.6,
 method="Tanimoto"), k=2 ,mode="b")
 byCluster(cl) 

## $`11`
## [1] "650011" "650092"
## 
## $`15`
## [1] "650015" "650069"
## 
## $`45`
## [1] "650046" "650054"
## 
## $`48`
## [1] "650049" "650050"
## 
## $`52`
## [1] "650059" "650060"
## 
## $`53`
## [1] "650061" "650062"
## 
## $`54`
## [1] "650063" "650064"
## 
## $`55`
## [1] "650065" "650066"
## 
## $`62`
## [1] "650074" "650085"

Output nearest neighbor table (matrix):

 nnm <- nearestNeighbors(fpset,numNbrs=6)
 nnm$names[1:4] 

## [1] "650001" "650002" "650003" "650004"

 nnm$ids[1:4,] 

## NULL

 nnm$similarities[1:4,] 

##     650001    650102    650072    650015    650094    650069
## sim      1 0.4719101 0.4288499 0.4275229 0.4247423 0.4187380
## sim      1 0.4343891 0.4246575 0.4216867 0.3939394 0.3922078
## sim      1 0.4152249 0.3619303 0.3610315 0.3424242 0.3367089
## sim      1 0.5791045 0.4973958 0.4192708 0.4166667 0.4104683

Trim nearest neighbor table:

 nnm <- trimNeighbors(nnm,cutoff=0.4) 
 nnm$similarities[1:4,]

##     650001    650102    650072    650015    650094    650069
## sim      1 0.4719101 0.4288499 0.4275229 0.4247423 0.4187380
## sim      1 0.4343891 0.4246575 0.4216867        NA        NA
## sim      1 0.4152249        NA        NA        NA        NA
## sim      1 0.5791045 0.4973958 0.4192708 0.4166667 0.4104683

Perform clustering on precomputed nearest neighbor table:

 jarvisPatrick(nnm, k=5,mode="b") 

## 650001 650002 650003 650004 650005 650006 650007 650008 650009 650010 650011 650012 650013 650014 
##      1      2      3      4      5      6      7      8      9     10     11     12     13     14 
## 650015 650016 650017 650019 650020 650021 650022 650023 650024 650025 650026 650027 650028 650029 
##     11     15     16     17     18     19     20     21     22     23     24     25     26     27 
## 650030 650031 650032 650033 650034 650035 650036 650037 650038 650039 650040 650041 650042 650043 
##     28     29     30     31     32     33     34     35     36     37     38     39     40     41 
## 650044 650045 650046 650047 650048 650049 650050 650052 650054 650056 650058 650059 650060 650061 
##     42     43     11     44     11     45     46     47     48     49     50     51     52     53 
## 650062 650063 650064 650065 650066 650067 650068 650069 650070 650071 650072 650073 650074 650075 
##     54     55     56     57     57     58     59     11     60     61     62     63     64     65 
## 650076 650077 650078 650079 650080 650081 650082 650083 650085 650086 650087 650088 650089 650090 
##     66     67     68     69     37     70     71     72     64     73     74     75     76     77 
## 650091 650092 650093 650094 650095 650096 650097 650098 650099 650100 650101 650102 650103 650104 
##     78     11     79     80     81     82     83     84     85     86     87      1     88     89 
## 650105 650106 
##     90     91

Using a user defined nearest neighbor matrix:

 nn <- matrix(c(1,2,2,1),2,2,dimnames=list(c('one','two'))) 
 nn

##     [,1] [,2]
## one    1    2
## two    2    1

 byCluster(jarvisPatrick(fromNNMatrix(nn),k=1)) 

## $`1`
## [1] "one" "two"

18.4 Multi-Dimensional Scaling (MDS)

To visualize and compare clustering results, the cluster.visualize function can be used. The function performs Multi-Dimensional Scaling (MDS) and visualizes the results in form of a scatter plot. It requires as input an APset, a clustering result from cmp.cluster, and a cutoff for the minimum cluster size to consider in the plot. To help determining a proper cutoff size, the cluster.sizestat function is provided to generate cluster size statistics.

MDS clustering and scatter plot:

 cluster.visualize(apset, clusters, size.cutoff=2, quiet = TRUE) # Color codes clusters with at least two members. 
 cluster.visualize(apset, clusters, quiet = TRUE) # Plots all items.

Create a 3D scatter plot of MDS result:

 library(scatterplot3d) 
 coord <- cluster.visualize(apset, clusters, size.cutoff=1, dimensions=3, quiet=TRUE) 
 scatterplot3d(coord) 

Interactive 3D scatter plot with Open GL (graphics not evaluated here):

 library(rgl) rgl.open(); offset <- 50;
 par3d(windowRect=c(offset, offset, 640+offset, 640+offset)) 
 rm(offset)
 rgl.clear() 
 rgl.viewpoint(theta=45, phi=30, fov=60, zoom=1)
 spheres3d(coord[,1], coord[,2], coord[,3], radius=0.03, color=coord[,4], alpha=1, shininess=20) 
 aspect3d(1, 1, 1) 
 axes3d(col='black')
 title3d("", "", "", "", "", col='black')
 bg3d("white") # To save a snapshot of the graph, one can use the command rgl.snapshot("test.png").

18.5 Clustering with Other Algorithms

ChemmineR allows the user to take advantage of the wide spectrum of clustering utilities available in R. An example on how to perform hierarchical clustering with the hclust function is given below.

Create atom pair distance matrix:

 dummy <- cmp.cluster(db=apset, cutoff=0, save.distances="distmat.rda", quiet=TRUE) 

## 
## sorting result...

 load("distmat.rda") 

Hierarchical clustering with hclust:

 hc <- hclust(as.dist(distmat), method="single") 
 hc[["labels"]] <- cid(apset) # Assign correct item labels 
 plot(as.dendrogram(hc), edgePar=list(col=4, lwd=2), horiz=T) 

Instead of atom pairs one can use PubChem’s fingerprints for clustering:

 simMA <- sapply(cid(fpset), function(x) fpSim(fpset[x], fpset, sorted=FALSE))
 hc <- hclust(as.dist(1-simMA), method="single") 
 plot(as.dendrogram(hc), edgePar=list(col=4, lwd=2), horiz=TRUE) 

Plot dendrogram with heatmap (here similarity matrix):

 library(gplots) 

## 
## Attaching package: 'gplots'
## 
## The following object is masked from 'package:stats':
## 
##     lowess

 heatmap.2(1-distmat, Rowv=as.dendrogram(hc), Colv=as.dendrogram(hc), 
              col=colorpanel(40, "darkblue", "yellow", "white"), 
              density.info="none", trace="none") 

19.1 Get Compounds from PubChem by Id

The function getIds accepts one or more numeric PubChem compound ids and downloads the corresponding compounds from PubChem Power User Gateway (PUG) returning results in an SDFset container. The ChemMine Tools web service is used as an intermediate, to translate queries from plain HTTP POST to a PUG SOAP query.

Fetch 2 compounds from PubChem:

 compounds <- getIds(c(111,123))
 compounds 

19.2 Search a SMILES Query in PubChem

The function searchString accepts one SMILES string (Simplified Molecular Input Line Entry Specification) and performs a >0.95 similarity PubChem fingerprint search, returning the hits in an SDFset container. The ChemMine Tools web service is used as an intermediate, to translate queries from plain HTTP POST to a PubChem Power User Gateway (PUG) query.

Search a SMILES string on PubChem:

 compounds <- searchString("CC(=O)OC1=CC=CC=C1C(=O)O") compounds 

19.3 Search an SDF Query in PubChem

The function searchSim performs a PubChem similarity search just like searchString, but accepts a query in an SDFset container. If the query contains more than one compound, only the first is searched.

Search an SDFset container on PubChem:

 data(sdfsample); 
 sdfset <- sdfsample[1] 
 compounds <- searchSim(sdfset) 
 compounds 

listCMTools()

##      Category                           Name      Input     Output
## 1      Upload                Upload CSV Data  character data.frame
## 2      Upload      Upload Tab Delimited Data  character data.frame
## 3  Properties             JoeLib Descriptors     SDFset data.frame
## 4  Properties          OpenBabel Descriptors     SDFset data.frame
## 5  Clustering             Binning Clustering     SDFset  character
## 6  Clustering Multidimensional Scaling (MDS)     SDFset  character
## 7  Clustering        Numeric Data Clustering     SDFset  character
## 8  Clustering        Hierarchical Clustering     SDFset  character
## 9      Search                  pubchemID2SDF data.frame     SDFset
## 10   Plotting               Graph Visualizer     igraph  character
## 11 Properties           ChemmineR Properties     SDFset data.frame
## 12  ChemmineR                  sdf.visualize     SDFset     SDFset
## 13     Search                      EI Search     SDFset    integer
## 14     Search             Fingerprint Search     SDFset    integer

toolDetails("Fingerprint Search")

## Category:        Search
## Name:            Fingerprint Search
## Input R Object:      SDFset
## Input mime type: chemical/x-mdl-sdfile
## Output R Object: integer
## Output mime type:    text/fp.search.result
## ###### BEGIN DESCRIPTION ######
## PubChem Fingerprint Search
## ####### END DESCRIPTION #######
## Option 1: 'Similarity Cutoff'
## Allowed Values:  '0.5' '0.6' '0.7' '0.8' '0.85' '0.9' '0.91' '0.92' '0.93' '0.94' '0.95' '0.96' '0.97' '0.98' '0.99'
## Option 2: 'Max Compounds Returned'
## Allowed Values:  '10' '50' '100' '200' '1000'
## Example function call:
##  job <- launchCMTool(
##      'Fingerprint Search',
##      SDFset,
##      'Similarity Cutoff'='0.5',
##      'Max Compounds Returned'='10'
##  )

20.1 Launch a Job

When a job is launched it returns a job token which refers to the running job on the UC Riverside cluster. You can check the status of a job or obtain the results as follows. If result is called on a job that is still running, it will loop internally until the job is completed, and then return the result.

Launch the tool pubchemID2SDF to obtain the structure for PubChem cid 2244:

job1 <- launchCMTool("pubchemID2SDF", 2244)
status(job1)
result1 <- result(job1)

Use the previous result to search PubChem for similar compounds:

job2 <- launchCMTool('Fingerprint Search', result1, 'Similarity Cutoff'=0.95, 'Max Compounds Returned'=200)
result2 <- result(job2)
job3 <- launchCMTool("pubchemID2SDF", result2)
result3 <- result(job3)

Compute OpenBabel descriptors for these search results:

job4 <- launchCMTool("OpenBabel Descriptors", result3)
result4 <- result(job4)
result4[1:10,] # show first 10 lines of result

##         cid abonds atoms bonds dbonds HBA1 HBA2 HBD   logP      MR       MW nF sbonds tbonds  TPSA
## 1      2244      6    21    21      2   12    4   1 1.3101 44.9003 180.1574  0     13      0 63.60
## 2      5161     12    29    30      2   15    5   2 2.3096 66.8248 258.2262  0     16      0 83.83
## 3     68484      6    24    24      2   14    4   0 1.3985 49.2205 194.1840  0     16      0 52.60
## 4     10745     12    34    35      3   18    6   1 2.5293 76.3008 300.2629  0     20      0 89.90
## 5    135269      6    30    30      2   18    4   1 2.4804 59.3213 222.2372  0     22      0 63.60
## 6     67252      6    22    22      1   13    3   1 1.7835 44.7003 166.1739  0     15      0 46.53
## 7    171511      6    25    23      2   16    5   2 1.2458 47.9481 222.4777  0     15      0 72.83
## 8   3053800      6    39    39      2   24    4   1 3.6507 73.7423 264.3169  0     31      0 63.60
## 9  71586929      6    38    33      2   29    7   6 1.7922 60.7157 294.2140  0     25      0 91.29
## 10    78094      6    24    24      2   14    4   1 1.6185 49.8663 194.1840  0     16      0 63.60

20.2 View Job Result in Browser

The function browseJob launches a web browser to view the results of a job online, just as if they had been run from the ChemMine Tools website itself. If you also want the result data within R, you must first call the result object from within R before calling browseJob. Once browseJob has been called on a job token, the results are no longer accessible within R.

If you have an account on ChemMine Tools and would like to save the web results from your job, you must first login to your account within the default web browser on your system before you launch browseJob. The job will then be assigned automatically to the currently logged in account.

View OpenBabel descriptors online:

browseJob(job4)

Perform binning clustering and visualize result online:

job5 <- launchCMTool("Binning Clustering", result3, 'Similarity Cutoff'=0.9)
browseJob(job5)