About

Deeplab (deeplab.ai) bridges the gap between state-of-the-art research and challenging industrial problems. Counts more than 18 researchers/engineers working in real-life projects while conducting high-end deep-learning research in large-scale product-level pipelines dealing with >1B data per day. Deeplab has successfully applied deep-learning in diverse and long term projects: 1) recommendation systems, including multiple modalities such as image, text, and other inputs 2) anomaly detection in fraud detection, 3) visual objects’ relations recognition and explainable AI/ML, as well as with 4) bioinformatics fields such as virtual screening for drug discovery, single cell multi-omics, cancer research all related to the state- of-the-art of deep learning. Recent success highlights include a self-funded project on COVID-19 drug discovery while being in the top-20 teams in the international JEDI grand challenge whose submissions are currently tested in vivo.

Solution & Product

Our focus is on structure-based virtual screening (SBVS), a process by which target-compound interactions are exploited with the aim of building models that can generalise to novel targets and compounds. Traditional approaches for SBVS employ physics-based molecular docking simulations, and are usually extremely slow and can therefore severely hinder the virtual screening throughput. In Deeplab, we have developed DeNViS, an AI-based virtual screening system using geometric deep learning, a sub-field of ML which is very powerful for processing graph data, i.e. proteins and chemical compounds. By leveraging the power of AI and avoiding physics-based simulations, we achieve state-of-the-art virtual screening performance with several orders of magnitude faster screening times, namely 10^4 times faster. This feature renders our approach suitable for virtually screening of vast compound databases, scaling up to billions of molecules per day.

Innovation & Scalability

DeepLab strives to address one of the world’s most complex problems: drug discovery. The technologies employed lie within subfields within AI, computational biology and biotechnology. We employ ML-related approaches involving geometric-DL and recent DL advances. By leveraging the power of AI we achieve state-of-the-art virtual screening performance with several orders of magnitude faster screening times, namely 10^4 times faster. This feature renders our approach suitable for scaling up to billions of molecules per day.Our algorithm is AI-based virtual screening system using geometric deep learning, a sub-field of ML which is very powerful for processing graph data, i.e. proteins and chemical compounds. Our business model is B2B based on custom drug discovery services on a project basis, tailoring our algorithmic product to the existing needs of our client. Our unique proposition is a VSaaS – Virtual Screening as a Service product. DeNViS can be used both as a tailor-made and also an “off the shelf” solution to run virtual screening when given specific drug targets and data libraries (either from client or acquired). We have the ability, agility and flexibility to sell our key differentiator: client-based high-performing/throughput solution based on innovative approaches. Our competitors sell just standalone products but we provide a service for faster virtual screening, suitable for massive databases with minimal requirements.

Traction & achievements

In JEDI Grand Challenge we virtually screened 1 Billion molecules against 3 proteins of Sars-Cov2. Currently our submission is evaluated in the wet-lab together with the top-20, out of ~138 teams.

Team & Why Us

Having a solid academic/research background (>15 years) in signal processing and machine learning, and a deep industrial experience during the recent 4 years, we are motivated enough to develop as well as to support actions and cultivate synergies with distinguished experts. Our team consists of: - Stavros Theodorakis: Founder, CEO - Vassilis Pitsikalis: Founder, CSO - Agamemnon Krassoulis: Senior Machine Learning Engineer - Stefanos Angelidis: Senior Machine Learning Engineer - Prodromos Kolyvakis: Senior Machine Learning Engineer - Thanos Tagaris: Senior Machine Learning Engineer - Petros Katsileros: Machine Learning Engineer - Alexandros Georgogiannis: Machine Learning Engineer - Nikos Antonopoulos: Machine Learning Engineer - Markos Diomataris- Machine Learning Engineer - Manos Zaranis: Machine Learning Engineer - Mary Parelli: Machine Learning Engineer - Afroditi Kravari: HR & Office Administrator - Stefanos Leptidis, Bioinformatician/ Bioinformatics lead, big-data application development - Alexandros Pittis: Bioinformatician

Funding Needs

To achieve our short to medium term goals, we are seeking seed funding to support our growth and goals regarding access to appropriate data, marketing, business development, development and hiring of experienced personnel. Specifically: - Team: Extend hiring of core team with executives, business, bioinformatics experts - Service: Develop/deliver product v2.0 to address issues/challenges far beyond the state-of-the-art extending v1.0 and advancing by a lot our competitive advantage - Pilot tests, collect feedback - Partner/Data: Partnerships and collaboration vendors/joint ventures to get more real-world data, have more disease specific evaluations to assist building our next version - Enrich service consultation to our customers - Future growth engineering: extra functionalities/innovations to expand addressable market.