In January our Protein A.I. start-up managed to achieve another important milestone. Our proprietary Protein Function Prediction A.I. model managed to achieve the highest accuracy and recall globally. Based on either an aminoacid sequence or a PDB (Protein Data Bank) file, our A.I. model is not only able to predict the function(s) of that protein very accurately (>99,8% accuracy) but also with a high recall (low number of false negatives).
Functions are predicted in the form of Enzyme Commission Numbers (EC) and Gene Ontology (GO) codes. Apart from predicting the function, our model is also able to exactly pinpoint the part of the protein responsible for that function. This function prediction feature is important for a number of reasons:
– In pharma drug development it is important to know which part of the protein causes a malicious activity to determine which part should be “blocked” by a medicine
– But also in Food and Chemistry environments you want to know which part of a protein causes a particular function and how you can change the behaviour of that particular part
– This function prediction component is also an important functionality in the area of personalized medicine (where a medicine could be personalized based on the DNA scan of a human being) -> although this is still future (5 to 10 years into the future probably)
Our whole Protein A.I. technology stack is now also ready for the commercial launch. Based on our own supercomputer capability, we now have a scalable software/hardware infrastructure that integrates the state-of-the-art A.I. models for all important building blocks that we need (folding, docking, function prediction and hallucination).
We already completed the first big test project successfully (replicating the results obtained in a lab environment with our computer-based A.I. models) and are running 3 more bigger tests right now across the domains of Chemistry, Microbiology and Pharma environments.
So if you are interested in running a joint innovation project with this new Protein A.I. start-up, please feel free to contact me.
During the development of this technology stack, we worked heavily together with both the ICT/AI departments at Fontys University and the AI/Data Science department at Maastricht University. And off course together with our partner at the Brightlands Chemelot Campus in Geleen.
#protein #functionprediction #artificialintelligence #genzai
