Cell Maps for AI

Bridge2AI project seeks to map the spatiotemporal architecture of human cells and use these maps towards the grand challenge of interpretable genotype-phenotype learning. In genomics and precision medicine, machine learning models are often “black boxes,” predicting phenotypes from genotypes without understanding the mechanisms by which such translation occurs. .

Teaming

The purpose of the Teaming module is to integrate and expand technical and scientific knowledge and expertise within a culture that promotes ethical considerations and diverse perspectives. The Teaming module will play a pivotal role in facilitating communication and collaboration among investigators and personnel from diverse geographical areas, disciplinary boundaries, and cultures.
Lead by:
PI: Jake Chen
CO-PI: Ying Ding

Data Acquisition

The Data Acquisition module will generate comprehensive datasets for protein targets of interest using three primary platforms: 1) Protein interaction mapping via affinity purification coupled to mass spectrometry to determine the interactome; 2) Spatial proteomics via imaging to map spatial subcellular coordinates; and 3) Genetic perturbation mapping to determine the functional states associated with the target via single-cell CRISPR-Cas perturbation screens.
Lead by:
PI: Emma Lundberg
Co-PI: Prashant Mali

Tools

The Tools module addresses three aims:

  • Dissemination of project data and cell maps.
  • Creation of tools for building cell maps from the three primary data streams.
  • Creation of tools that enable cell maps to power a range of AI/ML applications in the biomedical community.

All three of these aims will be supported by acquisition of Graphical Processing Unit (GPU) computing hardware, which will ramp up each year in anticipation of use in training and pilot use by ML researchers

Ethics

The Ethics module employs a rigorous systematic inquiry-driven methodology to tackle the unique and unsolved Ethical, Legal and Social Issues (ELSI) that CM4AI and its applications entail. To achieve this, we will use a value-sensitive design (VSD)-inspired approach to develop normative insights and address ethics and trust implications of cell maps-supported and visible ML-powered genomic medicine. Through unabated stakeholder engagement and a co-construction process, we will develop new ethics frameworks and practical tools for guiding data generation practices in a way that anticipates and addresses ethical challenges, from data generation and visible ML research and development, to clinical adoption and downstream health decisions and outcomes.
Lead by:
PI: Jean-Christophe Bélisle-Pipon, Vardit Ravitsky
Co-PI: Yael Bensoussan

Standards

The Standards module will provide a reusable, cloud-based informatics platform and tools for FAIR data and software interoperability standards implementation across the CM4AI Data Generation Project, based on the FAIRSCAPE digital commons framework.

Skills and Workforce

Alabdulkareem A,2018 Jul 18;4(7):eaao6030. doi: 10.1126/sciadv.aao6030. PMID: 30035214; PMCID: PMC6051733.
The Skills and Workforce Development module will bridge expertise across the CM4AI modules to prepare an expansive set of biomedical faculty, trainees and other NIH-funded scientists to ethically use the data sets, tools, and standards developed through the proposed project. The Module will 1) Develop and deliver extensible skills training components, 2) Create extensible workforce development components and work closely with Teaming and other modules to execute on the PEDP to recruit a diverse future workforce, and 3) Evaluate and iterate on training and workforce development components.