Compound Combination

When to use

• You optimized two positions of a series separately and want every compound that varies both at once
• You want a combinatorial library of the side groups that did well in your SAR, to dock or score
• You want to see which combinations pass molecular weight, cLogP or QED before deciding what to make

How it works

1. 01Find the core

   The common structure within each library is found first, and the part the two share becomes the core (RDKit maximum common substructure, MCS). Every position on the core that can carry a substituent gets a number.

2. 02Collect R groups

   Each compound is split against the core to see which group sits at each position. The groups at the position A varied become R1, the ones at the position B varied become R2, and each distinct group is counted once.

3. 03Join R1 × R2

   One of A's groups goes on the R1 position and one of B's on R2, for every pairing. Eight R1 groups and six R2 groups make 48 compounds.

4. 04Properties and tidy-up

   The joined structures are cleaned up, results that do not read as structures and duplicates are removed, and RDKit computes eight properties (MW, cLogP, TPSA, HBD, HBA, RotB, QED, SA).

Input formats

The file is read in your browser and never uploaded; the server receives the IDs and SMILES only, and names the IDs of any SMILES RDKit cannot read. Unlike the desktop version, no column has to be called smiles, and there is no property filter before submitting: you filter on the results page.

Tutorial: combine two aryl piperazine series

1-Acyl-4-aryl piperazines are a common medicinal chemistry scaffold. Series A keeps the acetyl group and varies the aryl ring on N4; series B keeps the phenyl ring and varies the acyl group on N1. Combining them puts each of A's 8 aryl rings with each of B's 6 acyl groups, 48 compounds in all. Run example at the top of the page fills in everything below.

1. 1

   Add the libraries

   In the Libraries step, drop a file onto A and onto B, or paste a list copied from a spreadsheet. The example is a list with an ID and a SMILES on each line, split by a tab.

2. 2

   Check the columns

   The wizard shows the SMILES and ID columns it found and previews the compounds with their structures on the right. If the drawings do not look like molecules, the wrong column was picked; change it in the column picker.

   
3. 3

   Pick the combination

   Once both libraries are ready, the Common Core step finds the shared structure on its own. Here it finds one core, O=C(N1CCN([*:1])CC1)[*:2], with "8 R1 from A × 6 R2 from B = 48 compounds" already picked. Above it, each library says how many of its compounds were used.

   
4. 4

   Review and submit

   Name the project in Project Settings and submit. The cost is 10 credits whatever the count.

5. 5

   Read the results

   On the results page, filter by property ranges and sort by any property. Each compound also names the library compounds its R1 and R2 came from.

   
6. 6

   Carry on

   Download the compound list and dock it against your target with Virtual Screening, or predict activity with a target model.

Reading the results

You may get fewer compounds than R1 × R2. Joined structures that do not read as structures are dropped, and two pairings that turn out to be the same compound are merged into one.

Example result: aryl piperazines, 8 × 6

A real job, shown read-only. All 48 pairings joined and none were duplicates. These are the five with the highest QED.

Cost and limits

A combination job costs 10 credits ($0.01 per credit) whatever its size. Even 50,000 combinations take a little over a minute to compute, and almost all of the elapsed time goes into bringing the compute up, so the price does not follow the number of compounds. Finding the common core is free. The wizard shows the quote before you submit and that is what is deducted.

A library holds up to 1,000 compounds and a job makes up to 50,000. A combination that would make more cannot be picked; use smaller libraries.

The example above (48 compounds) took about 70 seconds from submit to result, of which 67 seconds was waiting for an instance and the computation under 2 seconds, and cost 10 credits.

If the job fails, the credits come back in full. If you stop it yourself, half comes back.

Citations

When you use results from this workflow in a report or paper, please cite the methods below.

FAQ