Project 1: Knowledge

A significant amount of the pre-clinical data accumulated by BenevolentAI from the development of biological assays to characterise drug targets and evaluate potential therapeutic molecules. Databases dedicated to manage bioassay data could provide a rich resource for drug target identification. As such, the aim of this project is to build an automatic data processing pipeline to standardise assay data ingestion into BenevolentAI’s Knowledge Graph. The opportunities offered by this project include:  

• Learning and applying FAIR data principles (Findable, Accessible, Interoperable and Reusable)
• Inspecting assay data content features to understand how this information can support the Drug Discovery process
• Designing and implementing assay data models and data standardisation workflows
• Developing data engineering skills to support the integration of assay data into the BenevolentAI’s Knowledge Graph

Project 2: Target Identification

At BenevolentAI, our teams have developed performant models by collecting and encoding biological and chemical information around biomedical entities (e.g. genes, drugs, diseases, mechanisms). We are looking for an intern to help build a catalogue of discrete latent vector representations in our feature store; a central repository of features for biomedical entities in the problems we face day-to-day. This set of representations will help facilitate rapid experimentation of new models and improve existing workflows used for target identification. Due to the nature of the work, you will be expected to liaise with various teams across the business including chemistry, target identification, and precision medicine.  You will be supported by bioinformaticians, data scientists and engineers. The opportunities offered by this project include: 

• Using the existing data in our internal feature store to train and evaluate machine learning models
• Developing new features based on bio/chemoinformatic and machine learning approaches
• Supporting your team, you will implement classification models consuming biomedical data
• Performing benchmarking, validation, and evaluation of the created models and the features they use.

Project 3: Precision Medicine

The Precision Medicine team at Benevolent is applying latent variable models (LVMs) to transcriptomics data in order to define pathologically distinct patient subgroups and their key gene regulatory mechanisms. Within the Precision Medicine Omics pipeline, the latent variables are analysed in the context of a priori defined genesets, representing known biological processes or metabolic pathways. Given the availability of high resolution single cell transcriptomics data (scRNA-seq) for many human tissues and diseases, the team is planning to use these data and identify de-novo genesets, i.e. groups of genes representing disease-specific mechanisms that would provide an enriched view of the results of our AI models in the context of drug discovery. As an intern, you will be helping the Precision Medicine team to tackle this goal. You will be supported by bioinformatics / AI data scientists and will interact with engineers and Translational Medicine Scientists. This opportunities offered by this project include: 

• Exploring the data available at the Human Cell Atlas or analogous databases and select appropriate datasets for one disease of interest at Benevolent; 
• Running LVMs, adapting and/or developing other AI models to identify disease-specific mechanisms; 
• Collaborating with biology experts to analyse and evaluate the results. 
• Contributing to Python codebases to process and analyse the data