Structural and Functional Analyses of the Full-length Insulin Receptor (IR) and Type 1 Insulin-like Growth Factor Receptor (IGF1R) in the Liganded Active State

本次追加拨款申请是为了申请购买高性能GPU的资金 父项目下的单粒子Cryo-EM数据存储和处理集群 重点是对准确的配体诱导胰岛素激活机制的理解。 受体和IGF1R。在过去的一年里，我们在完成所有 3.拟建项目的目标。然而，在我提交了最初的提案之后，我们的冷冻-EM 设施得到显著扩展，目前拥有4台高端电子显微镜， 这是一年前的两倍。此外，我们的低温电磁设施最近 实现了一种新的波束-像移位数据采集方法，提高了数据采集的效率 数据收集吞吐量的另一个因子为3。因此，我们可以收集大约6 与一年前相比，某一时期的数据增加了一倍。因此，目前的两款老款GPU 我实验室的服务器不允许我们及时处理所有新生成的冷冻-EM数据 这将导致拟议项目的延误。此外，随着性能的提高 数据收集能力在我们的低温EM设施扩展后，我们可以产生4-5 TB的新低温- EM数据周刊。如此大量的数据无法在我的旧GPU服务器上长期存储。 通过使用具有16块功能强大的GPU卡和1000 TB存储空间的高级GPU集群，我 将能够在几天内完成一个完整数据集的处理。这也会让我尝试 几种新近发展起来的区分全长结构的三维分类方法 不同构象状态的受体/配体复合体，这将是这方面的关键结果 建议的项目。使用1000 TB的存储设备，我们可以存储50多种不同的 数据集在同一时间，这意味着每个数据集可以在GPU集群中存储~1 在转移到永久存储设备之前的一年。这将使我能够轻松地重新处理 这些数据改善了我们的结构，一旦一种新的图像处理方法被开发出来。

This application for supplemental grant is to request fund for purchasing a high performance GPU cluster for single particle cryo-EM data storage and processing, which is under the parent project that is focused on the understanding of the exact ligand induced activation mechanism of insulin receptor and IGF1R. In the last year, we have made tremendous progress toward finishing all the 3 aims of proposed project. Nevertheless, after I submitted the original proposal, our cryo-EM facility has been expanded significantly, currently housing 4 high-end electron microscopies, which is two times more than that one year ago. In addition, our cryo-EM facility has recently implemented a new data collection method with beam-image shift, which increases the throughput of data collection by another factor of 3. As a result, we can collect approximately 6 times more data in certain period, compared with one year ago. Thus, the current two old GPU servers in my lab would not allow us to process all newly generated cryo-EM data in a timely manner, which would lead to delay for the proposed project. Furthermore, with the improved data collection capability after our cryo-EM facility expands, we can generate 4-5 TB new cryo- EM data weekly. Such large amount of data cannot be stored in my old GPU servers for long term. By using an advanced GPU cluster that have 16 powerful GPU cards and 1000 TB storage space, I will be able to finish the processing of one full dataset in a few days. This will also allow me to try several recently developed 3D classification methods to distinguish the structures of full-length receptor/ligand complexes in different conformational states, which will be the key results in this proposed project. With the 1000 TB storage device, we could store more than 50 different datasets in the same time, meaning that each dataset can be stored in the GPU cluster for ~1 year before transferring to a permanent storage device. This would allow me to easily reprocess the data to improve our structures, as soon as a new method for image processing is developed.