mol-activity-annotator

csanno is a molecular annotator that uses ChEMBL to search to identify plausible targets. The question we are trying to solve in this program could be better defined with a simple example. Imagine we have one molecule and we want to identify probable targets in which it may be active. There are two possible situations:

1. The molecule is present in ChEMBL and therefore all the targets for which it was found active are retrieved
2. The molecule is not found in ChEMBL. And in that case what it will do is that it will look for similar molecules (using ChEMBL's similarity search) and will output the found targets and the number of molecule similars that were actually found active there

The tool also has a input file, option where for instance we might group a set of molecules with similar physiological activity, to try to identify common or most likely targets. All of this would be actually possible to do manually using ChEMBL online, but it would be tedious and extremely prone to errors.

csanno is heavily reliant on ChEMBL web services, therefore it is fundamental to have a good internet connection. Alternatively if a local ChEMBL installation is available it can be easily used with increased performance benefits.

The annotation profile of a molecule is essentially a report of the evidence that ChEMBL has for that specific structure or, if required, of the molecules that are very similar to it.

CSANNO requirements

CSANNO is essentially Python software that relies on a number of external libraries. Namely:

• numpy
• requests
• rdkit

Of those, rdkit is the trickiest to install. Anaconda makes the rdkit installation a breeze and this is the process recommended, although it might be possible to use a docker container with rdkit already installed. To install rdkit with anaconda we can follow the procedure detailed in the official rdkit distribution:

First we create an environment rdkit-env like this:

sh $ conda create -c conda-forge -n rdkit-env rdkit

and then to activate we just activate the environment to start working

sh $ conda activate rdkit-env

Windows users can omit the conda and just use activate rdkit-env at the windows command line prompt

With Anaconda, numpy is installed by default and installing the others can be performed simply with:

sh $ conda install requests

Just make sure these are installed within the rdkit environment.

Installing CSANNO

This should be a trivial process. Just copy the csanno.py files to a new folder in your woring directory:

$ mkdir qsartools $ cp [source code path]/*.py qsartools/. $ cd qsartools

SAR files

SAR files are the required format for most of the operations. SAR files are text files that should store activity data for a given target for a set of molecules and each line represents one molecule. Columns are separated by <tab>. The basic structur of a SAR file is

• The first column is an alphanumeric identifier for the molecule (should be unique);
• The second column contains the activity registered for that molecule (may be binary qualitative)
• The third column contains the molecule SMILES

Here is a sample SAR file, where the identifiers are the ChEMBL IDs

CHEMBL158973 0.5000 CN(C)CCSC(C)(C)C CHEMBL476516 0.1880 NCCc1ccc(Br)cc1 CHEMBL309689 0.0000 Oc1noc2c1CCNCC2 ... CHEMBL2331792 0.0000 COc1noc2c1CCNCC2 CHEMBL2331804 0.0000 Cn1oc2c(c1=O)CCNCC2 CHEMBL2436555 0.2200 NC1=Nc2ccc(Cl)cc2CN1

Asking for Help

csanno has a built in helpthat clarifies all its options

sh $ python csanno.py -help producing ``` CSANNO - (C) 2019/2023 - Andre O. Falcao DI/FCUL version 0.3.20230605 Usage: This is a Python3 tool and requires an enviroment where RDkit and requests are installed. To run type python csanno.py -in [input .sar file] [options]

Control parameters: -in file_name - the data set to annotate (required) (.sar format) -sim thr - similarity threshold [if absent (default) no similarity search will be performed] -mol [molecule in SMILES format] - molecule for which analysis is going to be performed. Incompatible with -in option -report [AD]: A - aggregate report; D - Detailed report (one line for each molecule) Both options simultaneusly are permitted and both reports are produced -search [ANU] A - searches for actives in the database (default); N - searches non-actives; U - searches unknowns (compounds for which activity is not determined) All options can be included simultaneously -ofield [GUO] G - outputs the unique gene (default); U - outputs the uniprot id; O - outputs the organism All options can be included simultaneously

Output Control Options: -joint_aggs - produces a report with conjoined similarities and exact matches -silent - no intermediate output at all -help - this screen -nofiles - no files are created -toscreen - writes output to screen -pickle - writes a Python pickle for easy posterior analysis (see internal documentation) ```

Essentially it allows to search for ChEMBL structural similars and ellaborating a report out of it

Molecular activity annotation

csanno is a molecular annotator that uses ChEMBL to search to identify plausible targets. The question we are trying to solve in this program could be better defined with a simple example. Imagine we have one molecule and we want to identify probable targets in which it may be active. There are two possible situations:

1. The molecule is present in ChEMBL and therefore all the targets for which it was found active are retrieved
2. The molecule is not found in ChEMBL. And in that case what it will do is that it will look for similar molecules (using ChEMBL's similarity search) and will output the found targets and the number of molecule similars that were actually found active there

The tool also has a input file, option where for instance we might group a set of molecules with similar physiological activity, to try to identify common or most likely targets. All of this would be actually possible to do manually using ChEMBL online, but it would be tedious and extremely prone to errors.

Simple single molecule annotation

The simplest way of running the csanno is by testing only one molecule that can function as an input

Using the -mol option writes the result directly to the screen. Therefore running the app with zolpidem (SMILES:CN(C)C(=O)Cc1c(nc2ccc(C)cn12)c3ccc(C)cc3)

sh $ python csanno.py -mol CN(C)C(=O)Cc1c(nc2ccc(C)cn12)c3ccc(C)cc3

will produce the following output:

TSPO 1 1.0000 SLCO1B3 1 1.0000 SLCO1B1 1 1.0000 LMNA 1 1.0000 GABRA3 1 1.0000 GABRA2 1 1.0000 GABRA1 1 1.0000

This result is a table which must be read this way. In the first column it is the gene symbol associated with a given assay for which the molecule has been proven active. It is important to notice that activity here means that a positive response has been found for it in at least one assay, whether it is Activation, Ki, IC50 or simple inhibition. If a positive result has been found, then it is reported. The next column refers to how many molecules were actually measured as active in that target. As we have only used one molecule, the only possible result is 1. Finally, the last column with values of 1.0000, refers to the ratio of the input molecules for which there were found actives. Again as we have tested only one molecule, this is the only possible output

Simple multiple molecule annotation

To test for the multiple targets we have created a simple file with 5 molecules, all known anti depressants to show the common binding profiles

Sertraline NA ClC1=CC=C([C@H]2C3=C([C@H](CC2)NC)C=CC=C3)C=C1Cl Citalopran NA Fc1ccc(cc1)C3(OCc2cc(C#N)ccc23)CCCN(C)C Fluoxetine NA CNCCC(c1ccccc1)Oc2ccc(cc2)C(F)(F)F Tradozone NA Clc4cccc(N3CCN(CCCN1/N=C2/C=C\C=C/N2C1=O)CC3)c4 Venlafaxine NA OC2(C(c1ccc(OC)cc1)CN(C)C)CCCCC2

Running the following command

sh $ python csanno.py -in anti-dep.sar -sim 0.7 -report AD

will produce 2 files. The name of those files is created based on the .sar file. So using anti-dep.sar will produce:

• anti-dep_anno_count_T0.txt - This is a Tier 0 file and will contain the absolute and relative frequencies of the molecules in the input file that appear to be active in specific targets
• anti-dep_anno_count_T1.txt - This is a Tier 1 file and will contain the the absolute and relative frequencies of the molecules that are similar to the molecules in the input file to each specific target in which they were found active

The tier 0 file thus refers to the known targets in single-gene essays that are know to bind to these proteins. anti-dep_anno_count_T0.txt will look something like this:

SLC6A4 5 1.0000 SLC6A3 4 0.8000 SLC6A2 4 0.8000 SIGMAR1 4 0.8000 HTR2C 4 0.8000 REP 3 0.6000 NET 3 0.6000 KCNH2 3 0.6000 HTR2B 3 0.6000

This file is actually 52 lines long but only the top 8 are shown. It shows which are the individual targets common in all of them. As we can see, SLC6A4 (the serotonin transporter) is common to all 5 molecules, followed by SLC6A3 and SLC6A2 the dopamine and noradrenaline transporter, respectively. Also the the Sigma1 Receptor and the Serotonin 2C receptor are known to be affected by 4 out of 5 molecules

On the other hand we can look at anti-dep_anno_count_T1.txt for potential targets, as these refer to the activity profiles of similar molecules (in this case 70% similar)

SLC6A4 64 0.6737 SLC6A2 29 0.3053 SLC6A3 24 0.2526 REP 8 0.0842 ADRA1A 8 0.0842 NET 7 0.0737 HTR2A 6 0.0632 SIGMAR1 4 0.0421 of the 183 molecules found, the top spots were essentially the same, but the ADRA1A Adrenergic A1 receptor and the NET, the norepinephrine transporter, appear with a high representativity. Troubling might be the presence of HTR2A, in 6% of all matches.

To have a better look at which of the base molecules are associated, we may run the above command again with the following option -report AD This will create two more files that stand for the individual targets each molecule was active on. The format is a standard 'transaction type' file with one line for each molecule, followed by the found targets, named similarly to the other two files created:

• anti-dep_annotations_T0.txt - binding targets to each molecule in the input file
• anti-dep_annotations_T1.txt - binding targets to each molecule similar to the molecules in the input file.

Therefore anti-dep_annotations_T0.txt with the above options will look like this (end of lines arranged for this report, in the original, only one line per molecule is produced):

Sertraline: ADRA2B CACNA1C KMT2A SIGMAR1 CYP3A4 ADRA2A HTR2B REP CYP2C19 CHRM1 KCNH2 CHRM4 ADRA2C MC5R CHRM5 HTR2A SLC6A3 HTR2C CYP2D6 ADRA1B TMEM97 SLC6A2 CHRM2 SLC6A4 Citalopran: TMEM97 ADRA1B SLC6A2 HRH1 KCNH2 ADRA1A NET SIGMAR1 SMN2 ADRA1D SLC6A3 HTR2C REP HTR2B SLC6A4 SLC22A1 Fluoxetine: ABCB1 ADRA2B CACNA1C KCNK9 SIGMAR1 NET HRH3 ACHE CYP3A4 ADRA2A SLCO1B1 HTR3A CYP2C19 CHRM1 CYP2C9 HRH1 KCNH2 LMNA KCNK2 CHRM5 CYP2B6 HTR6 CHRM3 HTR2A DRD2 SLC6A3 SLCO2B1 HTR2C CYP1A2 CYP2D6 CYP2C8 SLC6A2 SLCO1B3 SLC6A4 Tradozone: ADRA2C HTR1B ADRA2B SLC6A4 HRH1 ADRA1A SIGMAR1 HTR2A DRD3 ALB ADRA1D DRD2 FAAH HTR2C ADRA2A HTR2B HTR1A ADRA1B Venlafaxine: SLC6A2 ADRA1A NET LMNA SLC6A3 REP SLC6A4

Similarly anti-dep_annotations_T1.txt with the above options will look like this:

Sertraline: NFKB1 TMEM97 SLCO1B3 SLCO1B1 REP HTR2A KCNH2 CYP3A4 ADRA1B CHRM2 SLC6A2 NET NS1 MAPK1 SLC6A3 ADRA2A SIGMAR1 CYP1A2 FFP ADRA2C KMT2A TP53 MC5R CACNA1C HTR2B CYP2D6 CHRM4 SLC6A4 ADRA2B CHRM1 CHRM5 HTR2C CYP2C19 Citalopran: NFKB1 TMEM97 SLCO1B3 SLCO1B1 REP KCNH2 ADRA1B CYP3A4 SLC6A2 SLC22A1 ADRA1D ADRA1A NET MAPK1 SLC6A3 SIGMAR1 ALOX15 SMN2 CYP2C9 HTR2B SLC6A4 CHRM1 HRH1 LMNA HTR2C CYP2C19 Fluoxetine: CYP2B6 NFKB1 HTR3A CYP2C8 SLCO1B3 SLCO1B1 HTR2A KCNK9 KCNH2 DRD2 CYP3A4 KCNK2 SLC6A2 TSHR NET ADRA2A SLC6A3 SIGMAR1 SLCO2B1 CYP1A2 HTR6 HRH3 SMN2 CYP2C9 CACNA1C BLM CYP2D6 ACHE ABCB1 RORC SLC6A4 ADRA2B CHRM1 HRH1 LMNA CHRM3 CHRM5 HTR2C AMPC HPGD CYP2C19 Tradozone: FAAH SLCO1B3 SLCO1B1 REP HTR2A DRD2 ADRA1B SLC6A2 ADRA1D ADRA1A HTR1A ADRA2A SIGMAR1 ADRA2C HTR2B ADRA2B SLC6A4 HRH1 HTR2C HTR1B ALB DRD3 HLA-A Venlafaxine: SLC6A4 SLC6A2 LMNA ADRA1A NET SLC6A3 REP

Please note that these results will include targets that do not apear on Tier 0, meaning possible targets never before tested, but some overlaps may naturally occur.

As an example of a possible conclusion is that in some of the 70% similar molecules to Sertraline some of them were verified active on NET and CYP2C9, strongly suggesting that Sertraline itself, although not registered as active in these two targets, might actually be so. Other types of observations might be held for other molecules.