Exploring COCONUT Locally - Part 2

Hidden goodies in COCONUT: NP Fingerprints, Sugar-free Forms, NP-Likeness, and more

In Part 1, I showed how to download COCONUT, restore it from its MongoDB dump, then load it in a Jupyter Notebook via PyMongo, ready to be explored.

Now, I will show how to use PyMongo to query the extensive metadata and properties (the ‘hidden’ goodies) that have been calculated for all natural products (NPs) in COCONUT.

These goodies do not appear on COCONUT Web because, well, there are just too many of them to display.

According to Maria and coauthors:

“…the number of the computed properties is quite big (73 fields in each document corresponding to one unique NP), only a selected fraction of them is displayed on the COCONUT web interface.”

…hence, this post :)

NB: full credit to colleagues at the Steinbeck Group, in particular Maria Sorokina, Jonas Schaub, Kohulan Rajan, and Christoph Steinbeck for the hard work that went into what I’m about to show below.

What’s in COCONUT MongoDB?

Let’s start with a basic query of COCONUT using PyMongo to see the range of metadata and properties available. We will query uniqueNaturalProduct, the main collection containing unified and curated NPs.

Data in MongoDB are stored as key-value pairs within documents that are organised into collections. Below are all the keys for the first document. Note that keys may be different from document to document. The full list is available in Sorokina et al.

Natural Product Fingerprints: PubChem, Circular, Extended

Over time, more pre-calculated fingerprints for COCONUT’s NPs became available in COCONUT MongoDB, calculated using built-in CDK libraries. Here’s what they look like and how to download them.

Sugarfree Natural Products

Schaub et al. worked extensively on deglycosylation of NPs, including analysis of COCONUT glycosides.

In COCONUT, we can query which NPs have what kinds of sugars (e.g., linear, ring), and even download their aglycons (sugar-free forms) as SMILES.

Other goodies in COCONUT worth exploring (not in Web)

• Different NP-likeness scores, already pre-calculated e.g., npl_score, npl_noh_score and npl_sugar_score.
• Ertl FG fragments in SMILES and their frequency - ertlFunctionalFragments

Update: COCONUT in PubChem

COCONUT has recently been uploaded to PubChem. For now, the metadata I described in this post are not available on PubChem, but I guess more calculated properties and cross-linked data will be available soon.