vLife Release 6.8

November, 2022

vLife 6.8 introduces new and improved solutions focusing on data protection and privacy, natural language processing, and knowledge graphs for smart detection and comprehension of entities and processes leading to drug discovery and treatment. With the usage of reusable accelerators, the new solutions are more maintainable and scalable.

Protecting and maintaining an individual's privacy is a significant concern when collecting sensitive information from groups or organizations. 

  • Differential Privacy on Image Models offers some of the most stringent theoretical privacy guarantees. DP-image protects users' personal information in images from both human and AI adversaries. The DP-Image is formulated as an extended version of traditional differential privacy, considering the distance measurements between feature space vectors of images. Differential Privacy is achieved on images by adding noise to an image feature vector.
  • Encrypted Training on Medical Text Data demonstrates the training of a machine learning model on various medical data sources by ensuring data privacy via means of data encryption with the help of computed secret and private keys. 


Organizing unstructured data from various sources like social media, medical records, and business documents into a structured format and deriving insights from these can aid in processes involving drug discovery, among other tasks.

  • Generating Knowledge Graphs from Biomedical Literature ingests biomedical literature from Pubmed for any disease and extracts entities and their relations. This, in turn, can be used to generate Knowledge Graphs that can help understand Drug-Drug interactions and other relationships like Drug-Disease, and Drug-Treatments.
  • The Knowledge Graph Chatbot embodies a smart application powered by Amazon Lex and a web-based chatbot interface that will exploit the interconnected nature of the data within our knowledge graph to provide contextually relevant answers to any question.
  • Named Entity Recognition for Biomedical Text Mining can help annotate large-scale biomedical texts and create a directory that will enable researchers to develop a Biomedical Knowledge Graph for a holistic view of entities. This will allow users to create a better business model/research findings backed by data from unstructured data sources.
  • With Visual Question Answering in Medical Imaging, one can mirror real-world scenarios and derive answers from medical images. This application focuses on the clinically relevant task of Diabetic Macular Edema (DME) staging from fundus imaging. It allows the user to upload any fundus image and a question regarding any element of its description to obtain a near-accurate natural language answer.

The following accelerators were developed with the intent of streamlining the initial phase of drug discovery:

  • Predicting Aqueous Solubility of Compounds is a critical determinant of its success or failure as a drug candidate. Good aqueous solubility ensures a higher bioavailability of any drug.
  • Molecular Compound Toxicity Prediction efficiently and quickly tests whether a given chemical compound can disrupt processes in the human body that may lead to adverse health effects.
  • Thermodynamic Stability of Protein using Deep Learning to understand how the melting temperature of proteins identified with the help of a protein's amino acid can aid in designing stable proteins for specific uses
  • Multimodal Single Cell Integration measures the abundance of DNA, mRNA, and Proteins in single cells. The bone marrow stem cells develop into more mature blood cells, the information using which one can identify the flow of information transfer and tackle diseases at cellular levels.