Structural data science methods and software to study immunotherapeutic proteins

我们已经开发了一个高效的基于Web的交互式分子可视化和结构分析软件(ICn3D)[Wang等人。2020年，Wang等人。2022]通过与NCBI结构小组的初步合作。我们成功地应用该软件研究了病毒蛋白与细胞表面受体的结构和相互作用[Youkharibache等人]。2020年]。ICn3D软件现正成为一个协作研究平台，正如我们最近对SARS-CoV-2和其他贝塔冠状病毒的序列结构分析所表明的那样，我们在从SARS到MERS、OC43、HKU1、HKU4和MHV等冠状病毒的受体结合结构域/基序(RBD/RBM)超二级结构中确定了特定的序列结构微同源。2020]用于通过中和抗体或其他治疗分子进行靶向。在进行这一分析时，我们还提出了隐藏序列同源性的结构校正，证明了综合分析方法的价值，以改善结构。我们通过iCn3D中的F.A.I.R机制实现了创新的数据共享功能。事实上，我们不仅仅是数据共享，因为整个分析协议都嵌入到可共享的永久链接中，以实现重复性、可扩展性和协作研究。随着软件变得跨学科，它也成为了一个整合不同数据流的平台。软件开发本身正在演变为一个协作的、开源的中心，新的开发团队从内部和外部社区加入，并通过与内部和外部合作者共同组织的黑客松[https://www.iscb.org/ismb2020-program/ismb2020-hackathon]，共同接触到更广泛的开发人员社区。我的研究的基本基础是研究分子系统，特别是蛋白质的自缔合决定因素，正如它们在分子组织的几个水平上的结构对称性所揭示的那样[Youkharibache 2019；Youkharibache，Tran，and Abrol 2020]。我们正在开发的研究分子相互作用的软件和我们现在正在处理的应用程序开始捕捉到这一愿景，我们正在探索对称分析作为一种数据组织机制的初步实施，旨在开发基于分子相互作用知识的治疗学。例如，虽然抗体的重链和轻链对称性是众所周知的，但单个免疫球蛋白结构域本身由固有的伪对称原域组成[Youkharibache 2019]，这一特性基本上被忽视，可以为抗体工程，特别是纳米体开辟新的途径。同时，许多细胞表面蛋白受体，从T细胞到它们的靶细胞(TCRs、CD4、CD8、CD28、CTLA4、PD1、PDL1等)。由Ig结构域组成，通过寡聚的伪对称排列相互作用，揭示了蛋白质结构域结合的决定因素，特别是Ig结构域。我们正在组装一个以免疫球蛋白为中心的数据库，它将为设计新的基于免疫球蛋白的免疫受体和抑制剂提供宝贵的数据。免疫球蛋白结构域是迄今为止最常见的免疫组结构折叠，它的伪对称组装模式是理解和设计抑制剂和调节剂的宝贵指南。然而，细胞表面还有其他重要的折叠：GPCRs、MFS、SLCs等，它们被用作免疫细胞相互作用、代谢调节或病毒进入的受体。我们已经证明了广泛的多孔膜蛋白，包括GPCRs和SLCs，确实是通过伪对称组装机制形成的[Youkharibache，Tran和Abrol 2020]。其次，对于基于Ig的蛋白质来说，GPCRs代表了细胞表面/免疫组中最重要的分子支架子集，SLCs也是最重要的。我们早些时候已经确定，蛋白质结构域的假对称性存在于20%的已知结构中，但在整膜蛋白中发现了更高比例的准对称[Youkharibache，Tran和Abrol 2020]，我们现在看到更高比例的蛋白质分布在表层蛋白质组中，特别是在免疫细胞上。我们的对称性分析为我们提供了一个研究分子相互作用的解码框架，我们正在积极开发方法和数据库，以便能够基于这些想法设计新的基于Ig的受体作为抗癌治疗药物。基于灵活性分析的抗CD19和抗BCMA汽车的特征使我们能够支持正在进行的临床试验中的观察[Brudno等人。同时，我们观察到晶体中形成了CAR-T单链抗体的自发重排[PDBid：7JO8 Cheung等人。2020]由Ig域关联的准对称性调解[Youkharibache 2019]。我们目前正在开发一种算法来检测和表征蛋白质和蛋白质复合体的柔性部分，以研究蛋白质的折叠和展开、构象变化，最重要的是，对于我们的一些应用来说，将灵活性与其潜在的序列结构决定因素联系起来。我们还在开发一个带注释的免疫蛋白数据库，将所有包含免疫球蛋白结构域的已知结构重新组合在一起，以研究细胞间免疫突触核心的界面，主要涉及T细胞及其受体。

We have developed a highly efficient interactive web-based software for molecular visualization and structural analysis (iCn3D) [Wang et al. 2020, Wang et al. 2022] through an initial collaboration with the NCBI structure group. We successfully applied the software to study the structure and interactions of viral proteins with cell surface receptors [Youkharibache et al. 2020]. The iCn3D software is now becoming a collaborative research platform as demonstrated by our recent sequence-structure analysis of SARS-CoV-2 and other beta coronaviruses where we identified specific sequence-structure micro-homologies in receptor binding domains/motifs (RBD/RBM) supersecondary structures of coronaviruses from SARS to MERS, OC43, HKU1, HKU4, and MHV [Youkharibache et al. 2020] for targeting by neutralizing antibodies or other therapeutic molecules. While performing this analysis, we also proposed structure corrections that were hiding sequence homologies, demonstrating the value of an integrated analysis approach to improve structures. We have implemented an innovative data sharing capability through a F.A.I.R mechanism in iCn3D. In fact, we go further than data sharing, as entire analysis protocols are embedded in sharable permanent links for reproducibility, extensibility, and collaborative research. As the software becomes cross-disciplinary, it is also becoming a platform to integrate diverse data streams. Software development itself is evolving into a collaborative, open-source hub with new development groups joining in from both in the intramural and extramural community, and collectively reaching out to a broader developers' community through hackathons [https://www.iscb.org/ismb2020-program/ismb2020-hackathon], co-organized with intramural and extramural collaborators. The fundamental basis of my research has been the study of self-association determinants of molecular systems, especially proteins, as revealed by their structural symmetries at several levels of molecular organization [Youkharibache 2019; Youkharibache, Tran, and Abrol 2020]. The software we are developing to study molecular interactions and the applications we are now tackling are beginning to capture this vision and we are exploring the initial implementations of symmetry analysis as a data organizing mechanism, aiming at developing therapeutics based on molecular interactions knowledge. For example, while antibodies' heavy and light chain symmetries are well known, the individual Immunoglobulin domains consist themselves of intrinsically pseudo-symmetric protodomains [Youkharibache 2019], a property largely ignored that can open new routes to antibody engineering, especially nanobodies. At the same time, many of the cell surface protein receptors, from T-cells to their target cells (TCRs, CD4, CD8, CD28, CTLA4, PD1, PDL1, etc.) are composed of Ig domains interacting through oligomeric pseudo-symmetric arrangements revealing the determinants of protein domain association, and Ig domains in particular. We are assembling an Ig-centric database that will provide invaluable data to design new Ig-based immunoreceptors and inhibitors. The Ig-domain is by far the most common structural fold of the immunome, and its pseudo symmetric assembly patterns are an invaluable guide to understand and design inhibitors and modulators. There are, however, other important folds on cell surfaces: GPCRs, MFS, SLCs, etc. that are used as receptors for immune cell interactions, metabolic modulations, or for viral entry. We have demonstrated that a wide range of polytopic membrane proteins, including GPCRs and SLCs, are indeed formed through a pseudo-symmetric assembly mechanism [Youkharibache, Tran, and Abrol 2020]. Second, to Ig-based proteins, GPCRs represent the most important subset of molecular scaffolds in the cell surfaceome/immunome, and SLCs are also high up in the list. We had established earlier that protein domains' pseudo symmetries are found in 20% of known structures overall, yet quasi-symmetry is found in higher proportion in integral membrane proteins [Youkharibache, Tran, and Abrol 2020], and we are now seeing an even higher percentage across the proteins of the surfaceome, especially on immune cells. Our symmetry analysis gives us a decoding framework to study molecular interactions, and we are actively developing methods and databases that can enable the design of new Ig-based receptors as anti-cancer therapeutics based on these ideas. The characterization of anti-CD19 and anti-BCMA CARs based on flexibility analysis have enabled us to support observations in ongoing clinical trials [Brudno et al. 2020]; at the same time, we have observed the formation of a spontaneous rearrangement of a CAR-T scFv in a crystal [PDBid: 7JO8 Cheung et al. 2020] mediated by quasi-symmetry of Ig domains association [Youkharibache 2019]. We are currently developing an algorithm to detect and characterize flexible parts of proteins and protein complexes to study protein folding and unfolding, conformational changes, and, most importantly, for some of our applications to relate flexibility to their underlying sequence-structure determinants. We are also developing an annotated Immunoproteins database regrouping all known structures containing Immunoglobulin domains in interaction to study the interfaces at the heart of immune synapses between cells, and primarily involving T-cells and their receptors.