THE SCIENCE & PEOPLE BEHIND SIFTLINK
Connect what is known.
Explore what is possible.
Test what matters.
We bring computational discovery, scientific evidence and industrial judgement together to turn complex questions into clearer development decisions.
ONE CONNECTED APPROACH
From ingredient chemistry
to a practical next step.
AI helps us explore and prioritise. Experts assess the evidence, uncertainties and commercial relevance. Experiments test selected hypotheses.
Chemistry
Scientific literature
Existing data
AI modelling
+ expert judgement
Science · IP · Market contextRanked opportunities
Evidence priorities
Development plans
DISCOVER
Explore biological potential.
Virtual cell modelling, candidate screening, evidence synthesis and composition analysis help prioritise applications and hypotheses.
ASSESS
Understand the position.
Evaluate scientific support, regulatory and health-claim pathways, patent landscapes and the commercial context.
DEVELOP
Build the next evidence.
Translate findings into studies, experiments and development milestones suited to the opportunity.
OUR TEAM
Scientific depth.
Industry experience.
Expert judgement.
We are a team of PhD-level experts with industry experience in ingredient, food and beverage innovation.
Our expertise spans science, AI modelling, intellectual property and innovation management. We work across these disciplines to connect the technical question with the business decision.
Meet our teamLOOK AT THE WORK
Evidence should be specific.
Model predictions are hypotheses to assess. Project results are reported with their study context. Explore how this approach translated into a laboratory-tested fermentation protocol.
Read the wheat-bran caseSELECTED RESEARCH
Read the work.
Research by members of the team, spanning biological inference and ingredient evidence.
Community-reported ingredient benefits and clinical evidence
Gkountakos, Goulas & Kourmpetis · medRxiv. Not peer reviewed.
2013 / PEER REVIEWEDA large-scale evaluation of computational protein function prediction
Radivojac et al. · Nature Methods. Community assessment of protein-function methods.
2010 / PEER REVIEWEDBayesian Markov random fields for protein function prediction
Kourmpetis et al. · PLOS ONE. Network-based inference research.
The historical method studies concern protein-function prediction; they do not validate every current Siftlink application.
LET’S MOVE FORWARD
Let’s explore the scientific fit.
Tell us about your ingredient, target application or development question.