nir-vana innovation network

The SME spans a broad range of scientific and technical domains including the life sciences, medicine, computational biology, AI and machine-learning. The French company has already a collaborative agreement with a world-renowned medical university on a topic related to precision medicine and is looking to expand its activity and looking for new partners.

The company is developing an AI-Core which combines machine-learning (ML), deep-learning (DL) and knowledge-base systems (KBS) technology. The combination of these 3 technologies is rare and could answer several challenges in the following fields:

- In medicine, the company offers to provide decision-support and answers to the urgent need to avoid failed diagnosis or inappropriate patient treatments. World-scale efforts are devoted to precision medicine that aims at customizing healthcare for the individual based on contextually matched population characteristics. Progress towards precision medicine is currently spearheaded by biostatistics and epidemiology, as well as machine-learning and deep-learning. Precision-medicine is a highly complex endeavour as it aims at combining data from multiple sources that include sophisticated lab tests results, images from advanced scanners, as well as omics data. The company's technology can significantly contribute to the interpretation and translation of multi-modal and multi-scale data into concrete medical actions.

- In drug-discovery, the great challenge corresponds to find new treatments against pathologies such as cancer, diabetes, heart diseases or viral infections. AI-based technologies, in which the company can participate, include virtual-screening of new compounds, de-novo drug design, and drug re-purposing. State-of-the-art ML-based approaches to drug-discovery can be applied to all these tasks.

- In drug manufacturing as in many sectors of activity, quality by design (QbD) has increasingly gained worldwide attention. This initiative's aim is to promote consistent product quality and lower costs of development. The main tasks of QbD in drug manufacturing includes monitoring and improving process design and manufacturing cycles in both up-stream (e.g. bioreactors) and down-stream (e.g. chromatography) phases. The company's AI-core can intervene in both these upstream and downstream processes.

- In the medical-device sector, analysis tools such a mass-spectrometry (MS) plays a pivotal role in omics technology through which biomolecular structures, functions, and interactions are being discovered and monitored. However, omics data are inherently complex, and continuous improvements of analysis pipelines, notably through AI, are necessary. The SME's know-how in extracting useful biological information and producing actionable medical knowledge can play a pivotal role.

With these sectors in mind, the company considers necessary to integrate KBS technology with ML/DL. Indeed, ML/DL technology offers passive data-driven methods with no pro-active capabilities nor explicit domain knowledge. In this respect, ML/DL methods must be combined with KBS technology (AI-Core) in order to tackle the above-mentioned challenges.

The company is keen on sharing its know how in AI with health and drug-related companies, or IT companies, in order to create value in projects such as those mentioned above. The French SME will adapt its generic methodology and architecture to its partners' specific projects and thus participate in their conception and development under technical or research agreement.

Advantages & innovations

All applications involving AI require a close cooperation and mutual understanding between data scientists and domain experts. This issue can be that data scientists and medical domain experts may experience difficulties in quickly forming a well-posed problem and thus finding workable solutions. This state-of-affairs is particularly true in knowledge-intensive domains that characterize all health-related sectors. Indeed, these sectors present specific challenges due to the highly specialized sub-fields that can include elements of physics, physical chemistry, organic chemistry, biochemistry, biology, physiology and medicine. The main advantage of the French company lies in the fact that it addresses these challenges since its expertise overlaps the said scientific domains. Another advantage related to its transdisciplinarity, is its ability to incorporate knowledge of the aforementioned expert domains as parts of decision-support solutions in order to produce expressive and pro-active AI-systems. Examples of expressiveness and pro-activity in the target sectors are: - in medicine, it would correspond to questioning a medical doctor's decision and proposing alternative solutions better suited to a given patient. - in drug discovery, it would correspond to the autonomous exploration of data banks of an AI-agent during virtual drug screening. - in drug manufacturing, it would correspond to reminding an operator what physicochemical and biological conditions must be satisfied for a bacterial strain to develop in the bioreactor. - in mass-spectrometry (medical-device sector) it would correspond to transforming ML/DL outputs (e.g. class labels) into actionable medical knowledge. Also part of the added value to the company's approach is the fact that KBS technology can be carried-out in data-analytics itself in order to overcome repetitive aspects of data preprocessing, model selection, hyperparameter tuning, and model validation.

Stage of development

Under development/lab tested

Partner sought

The company is looking for international partners (public and private) that need to initiate or scale-up the application of AI in their core activities, under technical or research cooperation agreement. The targeted partners are: - Health-care institutions intending to develop AI-based decision-support systems at the points of care. For instance, first-line patient diagnosis, sorting and management, as well as developing a practical framework for the application of precision-medicine. The framework would be built on the partners' own patient databases as well as public databases in addition to evidence-based data derived from biostatistics and epidemiological studies. - Big pharma and SMEs involved in drug discovery: pharmacokinetic / pharmacodynamic PKPD (ADMET Absorption-Distribution-Metabolism-Excretion Toxicity, bioactivity…) prediction of new compounds, identification of new targets of the said compounds (virtual screening), QSAR (quantitative structure-activity relationship), discovering novel formulations by the virtual exploration of the chemical space, etc. - SMEs in the drug manufacturing industry: optimization of upstream (e.g. bioreactors) and downstream (e.g. chromatography) processes. In both cases, data science can establish meaningful correlations with input and control variables with variations in expression system results. - Medical devices companies: extraction of useful biological information and producing actionable medical knowledge from omics data. - SME with competencies in machine-learning, including cloud-based deployment and management of models in production (DevOps/MLOps), and front-end development (e.g. HTML, Java-script). If the front-end development is not part of the partner's core-competencies, then this component will be outsourced when and if needed. Ideally, the partner has a good experience in the application of machine-learning in the industry including DevOps/MLOps (cloud services, dockers, containers…), and wishes to apply ML to the aforementioned health-related and/or drug-related sectors. From the business standpoint, the partner will explore, with the French company, the value-creation and business opportunities with public or private enterprises and institutions. The partnerships can be technical cooperation agreements or research cooperation agreements depending on the project and the its degree of maturity.