Pharmaceutical companies spend billions of dollars on research and development to bring new drugs to the market. Chemists and Scientists dedicate their time to go through publications to study the latest in the field. Although the recent publications are annotated with machine-readable chemical descriptions (InChI), older publications cannot be scanned for chemical depictions. This makes it time consuming and cost intensive for the pharmaceutical companies.
By converting images back to their underlying chemical structure Molecular Translation can help (a) interpret old chemical images, (b) speed up R&D efforts by identifying and excluding previously published molecules from downstream drug discovery cycles and (c) identify novel trends by mining large datasets.

Join this session to learn how Quantiphi developed a custom modular solution for image captioning in the context of Molecular Translation. By implementing pure Tensorflow pipelines for both training and inference steps and making use of advanced optimizations we developed a cost-effective Molecular Translation inference model that is cheaper/faster by 60x to 200x than other state-of-the-art solutions.

https://www.nvidia.com/en-us/on-deman...