? PROJECT 3 The primary goal of this Program Project is to establish a quantitative and automated pipeline for the reconstruction and interpretation of highly complex cellular tomographic data. Here, we will focus on development of quantitative tools for tomogram annotation through deep learning and sub-tomogram averaging as well as interactive visualization tools. Specifically, in Project 3 we propose to address three aims:
Aim 1 : Adapt sub-tomogram alignment tools in EMAN2 for molecular template orientation determination and validation of localization results.
Aim 2 : Develop deep learning methods for semi-automated segmentation Aim 3: Incorporate pyCoAn and dependencies into EMAN2/SPARX to generate a single binary distribution and develop GUI interfaces for display and interactive tuning of results in the algorithms developed in all three projects. These efforts complement and feedback from the highly synergistic team of PIs both of Project 1 that will focus on the experimentally guided optimization of data collection strategies (Hanein) and on development and implementation of tomogram quality assessment and validation techniques (Penczek) and Project 2 (Volkmann), which focuses on automatic tomographic reconstruction technology, extraction of various features from the tomograms, and the analysis of distribution patterns derived from the extracted features.
