Help & Documentation
Learn how to search, explore, predict, and retrieve evidence in SCSMarineNP.
Introduction
The SCSMarineNP is a curated resource dedicated to marine natural products from Hainan Province and the South China Sea region.
This database provides information about marine-derived compounds, their source organisms, biological activities, chemical properties, and literature references. In addition to data searching and visualization, SCSMarineNP offers protein-based target screening and an evidence-grounded AI research assistant to support natural product discovery and literature exploration.
Searching the Database
Use the search bar on the homepage or navigation menu to quickly find compounds. Partial matching is supported, so you do not need to type the full compound name.
By Compound Name
Search by compound name or a keyword within the name.
HainanosideBy Source Organism
Search by genus, species, or organism name.
StylissaOn the Compounds page, use the filter panel to narrow results by:
Chemical Class
Filter by broad compound classes such as terpenoids, alkaloids, peptides, polyketides, phenolics, quinones, steroids, lipids, and glycosides.
Source Organism
Filter compounds by the marine organism from which they were isolated.
Molecular Weight
Use minimum and maximum molecular weight filters to focus on a specific chemical space.
Structure Search
Draw or input a SMILES structure to perform exact structure matching or substructure searching.
Sort & Page
Sort results by compound ID or molecular weight, and control the number of results displayed per page.
Browsing Data
Navigate through the main data sections from the top navigation bar.
Compounds
Browse marine natural products with 2D/3D structure viewers, molecular properties, Lipinski drug-likeness information, biological activity tables, and references.
Organisms
Explore source organisms with taxonomic classifications, collection information, and linked compound counts.
Target Screening
Submit one protein sequence to rank SCSMarineNP compounds by predicted activity probability.
AI Assistant
Ask questions about compounds, source organisms, bioactivities, scientific evidence, and website functions.
Compound Information
Each compound detail page provides the following fields when available:
| Field | Description |
|---|---|
| SCSMarineNP ID | Unique database identifier for each compound. |
| Compound Name | Common, systematic, or literature-reported compound name. |
| Molecular Formula | Elemental composition of the compound. |
| Molecular Weight | Molecular weight in Daltons. |
| SMILES | Canonical SMILES string for structure representation and cheminformatics use. |
| Chemical Class | Broad chemical class and original detailed class when available. |
| Source Organism | Marine organism from which the compound was isolated. |
| Collection Location | Reported geographic location or collection site. |
| Biological Activity | Reported activity type, target, value, unit, and reference. |
| Lipinski Properties Ro5 | MW, ALogP, HBD, and HBA values used for drug-likeness assessment. |
| 2D / 3D Structure | 2D structure image and 3D conformer viewer when available. |
| External Links | Links to public chemistry and natural product databases such as PubChem, ChemSpider, ChEMBL, COCONUT, NP Atlas, and CMNPD when identifiers are available. |
Compound detail pages include a dual-mode structure viewer:
2D Structure
Displays the compound structure as a 2D molecular image.
3D Conformer
Displays a computational 3D conformer generated from SMILES for visualization.
Organism Information
The Organisms page provides source organism information and taxonomic classifications. When NCBI Taxonomy IDs are available, taxonomic names can link to NCBI.
- Kingdom High-level biological kingdom Fungi
- Phylum Taxonomic phylum Ascomycota
- Class Taxonomic class Eurotiomycetes
- Order Taxonomic order Eurotiales
- Family Taxonomic family Aspergillaceae
- Genus Taxonomic genus Penicillium
- Species Species-level name when available Penicillium citrinum
Data Visualization
The Visualization page summarizes key database statistics and provides interactive charts for exploring the database.
Compound Classes
Shows the distribution of compounds by broad chemical class.
Physicochemical Properties
Displays distributions of molecular weight, ALogP, TPSA, and related properties.
Bioactivity Types
Summarizes the most common reported activity types.
Biological Sources
Summarizes source organisms by kingdom and phylum.
Target Screening
Target Screening predicts potential natural product–protein activity relationships. Given a single protein sequence, the module evaluates compounds in the SCSMarineNP library and ranks them by predicted activity probability. This is a sequence- and structure-based prediction workflow, not molecular docking or an experimentally measured binding assay.
The model uses ESM-2 to encode protein sequences and a Graph Isomorphism Network with Edge features (GINE) to represent small-molecule structures. Bidirectional residue–atom cross-attention integrates the two representations to model potential compound–protein interactions. The model is first pretrained on large-scale general compound–protein activity data, then fine-tuned on natural product-specific datasets through transfer learning.
Protein Encoding
ESM-2 derives sequence representations from amino acid residues.
Compound Encoding
GINE captures molecular graph features, including atoms and bonds.
Interaction Learning
Bidirectional cross-attention and transfer learning support natural product-focused predictions.
Enter a Protein Sequence
Open Target Screening in the navigation bar. In Protein input, upload a FASTA file or paste an amino acid sequence / FASTA entry. Submit one protein per job, with a maximum of 5,000 residues. You may use a FASTA header such as >my_protein to name the input.
Start Prediction
Click Start prediction. The Job status area reports progress and, when complete, summarizes the number of compounds evaluated and predicted active. Submitted jobs continue processing in the background if you close the tab or begin a new input.
Inspect Ranked Compounds
Review the Ranked compounds table, including rank, compound ID, name, structure, Predicted activity, and Probability. Click a compound ID to open its detail page.
Export the Results
Use Download full CSV to save the predicted compound rankings for further analysis. Select New input to prepare another protein sequence.
Probability (0–1)
The model's predicted activity probability; higher values indicate greater model-estimated likelihood of the active class, not calibrated certainty or measured binding affinity.
Active / Inactive
A binary label assigned using the model's saved classification threshold, which is displayed in the job results.
AI Assistant
The AI Assistant is an evidence-grounded research question-answering tool for SCSMarineNP. It helps users explore compound data, source organisms, biological activities, original publications, and website workflows through natural-language questions.
Open AI Assistant
Select AI Assistant from the navigation bar to open SCSMarineNP Research Copilot. The welcome screen provides sample questions for getting started.
Ask a Natural-Language Question
Enter a compound name or SCSMarineNP ID, ask for source or bioactivity information, request a literature summary, or ask how to use a website function. Click Send or press Enter; use Shift+Enter to insert a new line.
Review the Answer and References
Read the response and inspect the displayed data source or supporting references when available. Literature-supported responses may include DOI identifiers and PDF page numbers so findings can be checked against the original publication.
Continue or Start a New Chat
Ask follow-up questions in the same conversation, use suggested prompts when shown, or click New chat to start another conversation.
Compound Source
"What is the source organism of Lobophorin H?"
Literature Evidence
"Summarize the source, bioactivity, and original paper for Penicimarin F."
Protein Screening Guidance
"How do I use Target Screening with a FASTA sequence?"
Structure Search Guidance
"How do I perform a chemical structure search?"
The assistant routes incoming questions by intent. Compound identifiers and predefined query patterns can be recognized using rules, while a locally deployed Qwen2.5 model assists with less structured requests. Structured facts are retrieved from PostgreSQL through predefined, parameterized, read-only queries; the system does not execute model-generated SQL.
For supported compound-specific literature questions, linked publications are processed into page-annotated passages and indexed using BGE-M3 embeddings. Relevant text and curated database records can be provided to Qwen3.7-Plus through the Alibaba Cloud API to generate responses with supporting source information. When literature retrieval or the external language-model service is unavailable, the assistant can fall back to database-only answers.
Data Notes
3D Structures
3D conformers are generated computationally from SMILES for visualization and download. They are not experimental crystal structures.
Bioactivity Values
Bioactivity values are standardized when possible. Missing units may indicate qualitative descriptions or values not explicitly reported in the source record.
Chemical Classes
Compound classes are grouped into broad classes for browsing, while original detailed classes are retained on compound detail pages.
External Database Links
External links are shown when corresponding identifiers or structure-derived links are available.
Predicted Activity
Target Screening probabilities and binary labels are model outputs, not experimentally validated measurements. Interpret the Active / Inactive label using the threshold displayed for the job.
AI-Generated Responses
AI Assistant answers may rely on structured database records and/or available indexed publications. Review displayed references and check original sources before citing scientific conclusions.
Data Download
Data can be downloaded from compound detail pages and filtered compound result pages. Prediction rankings can also be exported from the Target Screening results page.
SMILES
Download canonical SMILES for the selected compound.
SDF
Download the available 3D SDF structure file.
CSV
Export filtered compound search results as a CSV file.
How to Cite
Please cite SCSMarineNP when using this database in research, reports, or publications.
FAQ
Contact Us
Your feedback helps us improve SCSMarineNP. We welcome reports on:
- Technical issues or bugs
- Feature requests and suggestions
- Data corrections or additions
- Collaboration opportunities
- Citation corrections