Salesforce, an organization most commonly recognized for Slack and cloud-based sales support software, sponsored the ProGen project last year, which used artificial intelligence to build proteins. The researchers behind ProGen, a research moonshot initiative, stated in their 2023 January website article that once made available for sale, it may help in the discovery of medical treatments that are more affordable than traditional approaches.
ProGen’s work resulted in a publication in the science magazine Nature Biotech demonstrating that artificial intelligence (AI) could effectively generate synthetic proteins’ three-dimensional architectures. Besides the publication, however, the effort was a failure at Salesforce and anywhere else, although not within the business perspective.
The fact is, as long as a short while ago.
Ali Madani, one of the scientists behind ProGen, founded Profluent with the hopes of transferring similar protein-generating technology from scientific research to the hands of pharmaceutical firms. In a TechCrunch interview, Madani explains that Profluent’s goal is to build “custom-fit” medicinal products by changing the traditional drug manufacturing model and beginning based on the requirements of treatment and needs of patients.
According to Madani, many medications are made of proteins, such as enzymes and antibodies. In the end, this is beneficial to patients who will be given medication in the form of a protein created by artificial intelligence.
While working in Salesforce’s research division, Madani became interested in the similarities between the “language” of proteins and natural language, such as English. Madani found that proteins may be thought of as words in a paragraph. Proteins are chains of bound amino acids that the human body relies on for many kinds of functions, from producing hormones to rebuilding the structure of muscle and bone tissue. Protein-related data can be utilized to anticipate entirely distinct proteins with unique functionalities when put into a generative artificial intelligence system.
Madani and Alexander Meeske, who founded together Profluent and a faculty member of microbiology at the University of Washington, hope to move forward with the idea by using it for the modification of genes.
According to Madani, ”proteins or enzymes taken directly from nature cannot cure many inherited disorders.” Genetic modification technologies that combine different strands of DNA to create new capabilities are also severely restricted by functional considerations. Profluent, on the other hand, may continuously optimize several characteristics to produce a [gene] operator that precisely meets the needs of every patient.
It is not completely unexpected. Other businesses and academic teams are demonstrating that artificial intelligence (AI) with generative capabilities is a useful tool for protein prediction.
MegaMolBART, a generative artificial intelligence (AI) algorithm developed by Nvidia in 2022, was developed on a dataset of several million molecules to predict chemical responses and find possible targets for medicines. Protein sequences were used to train a model named ESM-2, which Meta claimed was capable of predicting patterns of more than 600,000,000 proteins in a matter of two weeks. Additionally, AlphaFold, a system developed by Google’s AI research division DeepMind, anticipates whole protein structures with rapidity and precision that considerably surpass those of earlier, simpler algorithm techniques.
To develop novel modifying genes and protein-producing systems as well as enhance those that already exist, Profluent is developing artificial intelligence (AI) algorithms on enormous data sets, including datasets containing more than forty billion sequences of proteins. The business intends to work with other organizations to produce “genetic medicines” having the fastest potential paths for regulation instead of developing therapies on their own.
According to Madani, this strategy could significantly reduce the time and money normally needed for creating a cure. Industry association PhRMA estimates that the average time to produce a new medication, from discovery to regulatory approval, is ten to fifteen years. Meanwhile, current estimates place the price of creating a new medication among hundreds of millions of dollars to 2.8 billion dollars.
According to Madani, many significant medications were unintentionally found rather than purposefully created. The potential of Profluent presents an opportunity for humanity to progress from unintentional discovery to the deliberate creation of our most pressing biological solutions.
Profluent, a 20-individual startup situated in Berkeley, has the support of prominent venture capital firms such as Spark Capital, which led the business’s most recent $35 million investment circular along with Insight Partners, Air Street Capital, AIX Ventures, and Convergent Ventures. Jeff Dean, the head scientist at Google, has also made contributions, giving the platform a greater credibility.
Profluent’s priorities for the upcoming few months will be collaborator and consumer acquisition and improving its artificial intelligence models, which Madani believes would include boosting the data used for training. It will require to proceed quickly because competitors, such as EvolutionaryScale and Basecamp Research, are developing rapidly their models for producing proteins and raising substantial amounts of venture capital.
We’ve created our first platform and demonstrated scientific advancements in modifying genes, claimed the Madani. It’s high time to take action and begin facilitating ideas with collaborators who share our future vision.
