An AI workflow Khula built for Sawubona Mycelium — proof that the same approach that gives a professional their week back can compress weeks of lab research into hours.
Turn your mass spectrometry data into research-ready insights
A bespoke AI-powered metabolite biomining service for microbiological sourced extracts — built on a purpose-designed multistep algorithm that cleans, classifies, and researches your compounds automatically. Provided by Sawubona Mycelium.
One method. Many domains.
Khula Platform helps people reclaim hours every week by reworking the repetitive parts of their work with AI. This case study is that same promise, taken into a specialised scientific field.
For a fermentation-biotech, the slow, repetitive work isn’t admin — it’s manually researching hundreds of compounds from raw instrument data, one database at a time, for weeks on end.
Khula built a bespoke workflow that does it automatically: weeks of metabolite research compressed into a single, cited workbook. Different industry, same outcome — intelligence that scales, growth that lasts.
Weeks → hours
Manual literature research replaced by an automated, repeatable pipeline.
Admin to advanced science
The same Khula method, applied from everyday workflows to cheminformatics.
Built around real work
Shaped with the client’s own data, instruments, and quality standards.
Instrument
HPLC · GC-MS · LC-MS
Raw data
Hundreds of signals
Bespoke AI algorithm
Clean · Filter · Classify · Rank · Biomine
End-to-end report
Cited & scored workbook
The challenge with raw mass spectrometry data
When a researcher or scientist run microbiological sourced extracts through a mass spectrometer, the instrument produces a large spreadsheet of detected signals — often hundreds or thousands of rows. Most of those rows are noise, duplicates, or low-confidence detections. The compounds that matter are buried in the data.
Turning that raw output into something useful — knowing which compounds are real, what they are, where they come from, and what they are used for — normally takes weeks of manual database searching, literature review, and quality checking. Our service does this automatically, using a bespoke AI-driven multistep algorithm designed specifically for this type of exploratory biomining and data analysis.
The service is compatible with standard mass spectrometry data processing software outputs, accepting the Excel files your existing instruments and workflows already produce.
Hundreds of signals, only some are real
Raw data is full of noise. Our algorithm filters it down to only the compounds that meet strict identification standards.
Manual research takes weeks
Searching databases for each compound one by one is slow, error-prone, and hard to repeat. We automate the entire biomining process.
Results that are hard to trust
Without a clear confidence rating for each compound, it is hard to know which findings to act on. We score every identification.
A structured, repeatable multistep process
From raw data file to finished research workbook — every step automated and auditable.
You send us your data file
Share the Excel file from your mass spectrometry software, with any focus preferences you have.
Data cleaning and filtering
Rows failing quality standards are removed. Only detections passing every check move forward.
Compound classification
Each passing compound is classified as a natural plant product, external contaminant, or uncertain.
Ranking and biomining
Top-ranked compounds — up to 100 per run — are biomined across multiple scientific databases and published literature.
Report delivered
A formatted Excel workbook with five structured sheets. Everything cited from authoritative sources.
Five structured sheets. One complete workbook.
An end-to-end comprehensive report delivered as a structured Excel workbook.
The workbook bundles cleaned compound data, research, instrument settings, charts, and a full audit trail.
Cleaned and classified compound list
Every compound that passed quality filtering, with molecular formula, structural identifiers, mass accuracy, identification score, classification, and confidence rank. Duplicates removed, one row per unique compound.
Detailed compound research
For each top-ranked compound: chemical family, natural occurrence, discovery history, commercial uses, and relevant industries — biomined from PubChem, HMDB, ClassyFire, and MassBank.
Instrument settings record
The original acquisition and processing parameters from your mass spectrometry software, preserved in a dedicated sheet for traceability and reproducibility.
Research summary with charts
Aggregated statistics on chemical group distribution, industry relevance, and confidence scores — with embedded charts for immediate visual insight into what your extract contains.
Full run metadata and audit trail
A complete record of every database queried, every threshold applied, every source consulted, and deduplication statistics. You can see exactly how the analysis was run and reproduce it.
Everything in one file
All five sheets in a single formatted Excel workbook — no scattered files, no manual assembly. One file, everything cited, scored, and ready for publication, formulation, or regulatory documentation.
How the algorithm works
A purpose-built seven-stage pipeline handles everything from your raw data file to the finished workbook — automatically, transparently, and with a full audit trail at every step.
Receive inputs
Your data file is checked and accepted for processing.
Clean and filter
Signals that do not meet identification standards are removed.
Classify compounds
Each retained compound is classified as a natural product, xenobiotic, or uncertain.
Rank compounds
Compounds are ranked by a composite confidence score.
Quality checkpoint
A scientist reviews the ranked list before biomining begins.
Biomine databases
Top compounds are researched across six authoritative scientific databases.
Score and report
Every identification receives a 0–100 confidence score. The final workbook is assembled with full citations and audit trail.
Databases cross-referenced in priority order
What the algorithm will not do
No fabricated citations. No vendor-site sources. No Wikipedia. Compounds with low confidence scores are explicitly flagged as uncertain rather than reported as confirmed identifications.
From 400+ signals to 55 confirmed compounds
Two mycelium extract samples were processed through the BRAP Mycobiomolecule EDA pipeline. The before-and-after below illustrates the transformation from raw instrument output to a structured, cited research workbook.
400+ detected signals per sample. Most are noise, background, or duplicates — not immediately actionable.
| Signal ID | m/z | Score | Status |
|---|---|---|---|
| SIG-001 | 182.08 | 0.31 | Noise |
| SIG-002 | 251.10 | 0.41 | Low conf. |
| SIG-003 | 251.11 | 0.43 | Duplicate |
| SIG-004 | 309.22 | 0.38 | Low conf. |
| SIG-005 | 252.09 | 0.62 | Uncertain |
| SIG-006 | 183.09 | 0.28 | Noise |
55 confirmed compounds retained — classified, scored, cross-referenced, and fully cited.
| Compound | Group | Conf. | Industry |
|---|---|---|---|
| Cordycepin | Nucleoside | 94 | Pharma |
| Uridine | Nucleoside | 91 | Nutrac. |
| Ergosterol | Steroid | 87 | Cosmetics |
| Crotonoside | Nucleoside | 83 | Pharma |
| Oleic acid | Lipid | 79 | F&B |
Water Extract profile
Enriched in alkaloids and nucleosides (cordycepin, crotonoside, uridine). Pharmaceutical relevance: 46%.
Diluted Filtrate profile
Greater phenolic and terpenoid diversity, expanded iridoid glycosides. Nutraceutical relevance: 29%.
Representative analysis. Compound names and figures are generalised for illustrative purposes.
Built for researchers and teams working with mushroom, fungal, and microbial data
Academic and university researchers
If you are studying the chemistry of mushrooms, fungi, or microbiological natural product leads — including partners at NMU and SMU — BRAP turns your mass spectrometry data into a structured literature review, saving weeks of manual work and producing a comprehensive, database-backed metabolite inventory.
Pharmaceutical and biotech companies
If you screen fungal or fermentation-derived extracts for active compounds as part of early drug discovery, BRAP helps you quickly see what metabolites are present, how confident the identifications are, and what is already known about each compound in the published literature.
Nutraceutical and supplement brands
If you need to know the mycobiomolecule profile of your mushroom or fermentation-derived ingredients — for formulation, labelling, or regulatory documentation — BRAP gives you a clear, database-backed metabolite inventory with confidence scores for every detection.
Cosmetic and personal care formulators
If you work with mycelium or mushroom-derived extracts and need to understand their bioactive compounds, BRAP connects your raw mass spectrometry data to published science about each ingredient’s properties and known uses — supporting claims, formulation, and ingredient transparency.
Real-world scenarios
Screening a medicinal mushroom extract
A university research team studying a mushroom used in traditional African or Asian medicine gets over 400 detected signals from their mass spectrometer. BRAP returns a cleaned list of confirmed metabolites — classified, cross-checked, and linked to published literature. Within hours they know which ergosterols, polysaccharides, terpenoids, or phenolics are present and what the science says about each.
Verifying what is in a fermentation-derived ingredient
A nutraceutical company profiles a fermentation-derived mushroom extract before formulation. The output tells them whether expected bioactive metabolites are present, whether unexpected compounds showed up, and how confident each identification is — a science-backed basis for ingredient decisions and product documentation.
Finding lead compounds in a fungal extract library
A biotech company screens a library of fungal and mycelium extract samples. BRAP quickly profiles which samples contain high-confidence mycobiomolecule identifications — terpenoids, polyketides, phenolics, beta-glucans — and what is already known, helping prioritise candidates before committing to expensive bioassay work.
The team behind the service
Sawubona Mycelium (Pty) Ltd was founded in 2018 by Busi Moloi (CEO, MSc, MBA) and Neo Moloi in Centurion, Gauteng, South Africa. The name Sawubona comes from the Zulu greeting meaning “I see you” — a recognition of the unseen value within Africa’s extraordinary biodiversity.
We operate a fermentation-based bioprocessing platform producing bio-based ingredients from mushroom mycelium, with access to more than 3,000 fungi strains native to Southern Africa. Our work spans cosmetics, pharmaceuticals, nutraceuticals, functional foods, and beverages — now deepened by AI-powered bioinformatics tools like Mycobiomolecule EDA.
Key milestones
2018 — Company founded, combining expertise in fermentation science and medicinal mushrooms.
2019 — Joined the CSIR Biomanufacturing Industrial Development Centre (BIDC) programme.
2021–22 — First in Africa: 800 litres of liquid cultivated Enokitake mushrooms for bio-based cosmetics, in partnership with the CSIR.
2022 — Launched Blu Beryl — South Africa’s first fermentation-derived beta-glucan skincare products, supported by TIA, DSI, Innovation Hub and SEDA.
2023 — Exhibited at BIO2023 (Boston) and SynBioBeta2023, showcasing mycelium-derived bioactives internationally.
2025–26 — AI-powered Metabolite Biomining, including mushroom metabolite Effect Directed Analysis (EDA), to mine fermentation data for high-value biomolecules at scale.
Centurion, Gauteng, South Africa
We build it with you, not just for you.
Khula isn’t a faceless tool dropped in your lap. We sit beside the people who know the science — learning the workflow, shaping the algorithm around real data, and refining it together until it’s something the team trusts. This workflow was built hand-in-hand with the scientists at Sawubona Mycelium, the South African fermentation-biotech behind the BRAP service.
The results of this service rest on a bespoke AI-driven algorithm built specifically for exploratory data analysis and biomining of mass spectrometry data. It was purpose-designed for the task of cleaning, classifying, ranking, and researching plant metabolite data — using a multistep approach that combines domain-specific quality thresholds with automated literature mining.
General data analysis tools are flexible but not specialised. Our algorithm understands the structure of mass spectrometry output, distinguishes between natural metabolites and environmental contaminants, applies domain-specific quality thresholds, and is connected directly to the databases that matter for plant chemistry. It is built to do one thing well — and it does it with the precision and transparency that scientific work demands.