Inventa: A computational tool to discover structural novelty in natural extracts libraries

Abstract

Collections of natural extracts hold potential for the discovery of novel natural products with original modes of action. The prioritization of extracts from collections remains challenging due to the lack of a workflow that combines multiple-source information to facilitate the data interpretation. Results from different analytical techniques and literature reports need to be organized, processed, and interpreted to enable optimal decision-making for extracts prioritization. Here, we introduce Inventa, a computational tool that highlights the structural novelty potential within extracts, considering untargeted mass spectrometry data, spectral annotation, and literature reports. Based on this information, Inventa calculates multiple scores that inform their structural potential. Thus, Inventa has the potential to accelerate new natural products discovery. Inventa was applied to a set of plants from the Celastraceae family as a proof of concept. The Pristimera indica (Willd.) A.C.Sm roots extract was highlighted as a promising source of potentially novel compounds. Its phytochemical investigation resulted in the isolation and de novo characterization of thirteen new dihydro-β-agarofuran sesquiterpenes, five of them presenting a new 9-oxodihydro-β-agarofuran base scaffold.

Keywords: bioinformatic tools; computational metabolomics; mass spectrometry; natural products; prioritization; structural novelty discovery.

FIGURE 1

A conceptual overview of Inventa’s priority score and its components. (A) Feature Component (FC): is a ratio of the number of specific and unannotated features over the total number of features by extract. (B) Literature Component (LC): is a score based on the number of compounds reported in the literature for the taxon. It is independent of the spectral data. (C) Class Component (CC): indicates if an unreported chemical class is detected in each extract compared to those reported in the species and the genus. (D) Similarity Component (SC): Compares extracts based on their general MS² spectral information though their MEMO vectors and automatic outlier detectors. This score is independent from any retention-time based alignment procedure and complementary to FC. (E) The Priority Score (PS) is the addition of the four components. A modulating factor (w _n ) gives each component a relative weight according to the user’s preferences. The higher the value, the higher the rank of the extract. (F) Results Table is a resume of individual calculation components and results.

FIGURE 2

(A) UHPLC-HRMS chromatogram (BPI positive ion mode) showing the region where the dihydro-β-agarofuran sesquiterpenoids derivatives are suspected and displaying the only two compounds annotated for P. indica roots (plant with the highest PS). (B) Ion identity networking-based interactive ion map showing the combined results of the FC and CC for the IIN. In such display all features of a single neutral molecule are grouped under a single spot. The IIN are displayed according to their status (specific unannotated (blue), specific annotated (green), and non-specific unannotated -not interesting- (yellow)). Complementary information (adducts, row id, chemical class, etc.) are displayed interactively for each IIN if available, as shown in the zoom sections for the ion identity network 1734. The intensities in both cases (bar’s height and bubble’s size) are proportional to the original quantification table (before any filtering step). The scatter plot shows the m/z ratio of each feature (or ion network identity) on the y-axis. The feature-based ion map can be found in Supplementary Figure S82.

FIGURE 3

Original dihydro-β-agarofuran derivatives isolated from the Pristimera indica roots extract.

FIGURE 4

The relative position of the isolated compound (1–13) in the chromatogram for the ethyl acetate extract of Pristimera indica roots. The upper chromatographic trace corresponds to the ESI in positive ionization mode, while the lower trace corresponds to the Charged Aerosol Detector (CAD), a semi-quantitative trace. Compounds highlighted in green hold a new 9-oxodihydro-β-agarofuran base scaffold.

FIGURE 5

(A) HMBC Key and ROESY correlations for the compounds isolated from P. indica roots extract. (B) Experimental and B3LYP/def2svp//B3LYP/6-31G(d,p) calculated spectra in acetonitrile for compound 3.