Therapeutic Targeting a Non-Hodgkin Lymphoma Driver Using AI

使用AI靶向治疗非霍奇金淋巴瘤驱动因素 项目摘要 贝勒医学院（Baylor College of Medicine）和Atomwise Incorporation合作发现，优化和测试 使用人工智能（AI）来抑制不可药物化的癌蛋白。霍奇金淋巴瘤和非霍奇金淋巴瘤 霍奇金淋巴瘤（NHL）是从淋巴细胞开始的癌症，淋巴细胞是人体免疫系统的一部分。 霍奇金淋巴瘤和NHL之间的主要区别在于各自涉及的特定淋巴细胞： 如果存在称为Reed-Sternberg细胞的异常细胞，则淋巴瘤被归类为霍奇金淋巴瘤;否则，它 被归类为NHL。NHL是最常见的血液癌症，在美国每年导致超过20，000人死亡。 美国的有大约90种类型的NHL，通常发生突变时，在一个细胞内， 淋巴细胞MYD 88基因编码髓样分化初级反应88蛋白，其是一种关键的通用蛋白。 在炎症和免疫中具有重要功能的适配器。在刺激Toll样受体后， MYD 88转导信号以激活负责先天性和适应性免疫应答的基因。 MYD 88是一种在B细胞NHL中经常突变的驱动癌基因。最常见的误解 突变是L265 P，其将位置265处的亮氨酸改变为脯氨酸，并且占所有MYD 88的~ 90 突变。MYD 88 L265 P存在于约90%的Waldenström巨球蛋白血症（WM，一种罕见的NHL）中， 原发性淋巴结淋巴瘤和约29%的活化B细胞弥漫性大B细胞淋巴瘤（DLBCL）。WM是 被认为是不治之症DLBCL在约40%的患者中可以治愈，但那些MYD 88 L265 P的患者 比那些没有的人生存更差。Atomwise是第一个深度学习AI的发明者。 基于神经网络的技术和人工智能辅助药物发现的领导者，将对270万人进行虚拟筛选 化合物.我们鉴定了靶向MYD 88中L265 P附近的结合位点的AI选择的化合物的分数。我们 通过评估它们对MYD 88 L265 P泛素化和异种移植肿瘤发生的抑制来验证这些命中。 一种化合物用MYD 88 L265 P减弱NHL细胞的淋巴瘤生长，但用WT MYD 88则没有。 我们假设衔接子癌蛋白如MYD 88 L265 P可以被AI靶向。在这 申请，我们提出了两个具体的目标，以开发靶向MYD 88 L265 P的NHL候选药物 疗法在Aim 1中，我们将使用AI虚拟筛选数十亿种化合物，以发现新的候选药物 靶向MYD 88中L265 P附近的结合位点。在目标2中，我们将优化经验证的命中化合物， MYD88 L265P.从这种伙伴关系产生的数据将提供一个坚实的科学平台， 靶向致癌MYD 88 L265 P的发展，同时保留WT MYD 88，这对先天性和非先天性MYD 88的产生至关重要。 和适应性免疫。这项工作解决了直接靶向MYD 88 L265 P的未满足的临床需求， 推进针对突变特异性驱动因素的药物开发。

Therapeutic Targeting a Non-Hodgkin Lymphoma Driver using AI PROJECT SUMMARY Baylor College of Medicine (BCM) and Atomwise Incorporation have partnered to discover, optimize, and test inhibitors to undruggable oncoproteins using artificial intelligence (AI). Both Hodgkin lymphoma and non- Hodgkin lymphoma (NHL) are cancers that start in lymphocytes, which are part of the body’s immune system. The main difference between Hodgkin lymphoma and NHL is in the specific lymphocyte each involves: in the presence of abnormal cells called Reed-Sternberg cells, the lymphoma is classified as Hodgkin’s; otherwise, it is classified as NHL. NHL is the most common blood cancer and causes over 20,000 deaths every year in the United States. There are about 90 types of NHL, which usually develop when mutations occur within a lymphocyte. The gene MYD88 encodes myeloid differentiation primary response 88 protein, a critical universal adapter with essential functions in inflammation and immunity. Following stimulation of toll-like receptors, MYD88 transduces the signal to activate genes responsible for innate and adaptive immune responses. MYD88 is a driver oncogene that is frequently mutated in B-cell NHLs. The most frequent missense mutation is L265P, which changes leucine at position 265 to proline and accounts for ~90% of all MYD88 mutations. MYD88 L265P is found in ~90% of Waldenström macroglobulinemia (WM, a rare NHL), >50% of primary extranodal lymphomas, and ~29% of activated B-cell diffuse large B-cell lymphomas (DLBCL). WM is considered incurable. DLBCL can be cured in about 40% of the patients, but those with MYD88 L265P have poorer survival than those without. BCM collaborates with Atomwise, the inventor of the first deep learning AI technology based on neural networks and a leader in AI-assisted drug discovery, to virtually screen 2.7 million compounds. We identified scores of AI-selected compounds targeting a binding site near L265P in MYD88. We validated these hits by evaluating their inhibition of MYD88 L265P ubiquitination and xenograft tumorigenesis. One compound attenuated lymphoma growth from NHL cells with MYD88 L265P, but not that with WT MYD88. We hypothesize that adaptor oncoproteins such as MYD88 L265P can be targeted by AI. In this application, we propose two specific aims to develop drug candidates that target MYD88 L265P for NHL therapy. In Aim 1, we will use AI to virtually screen billions of compounds to discover novel drug candidates targeting a binding site near L265P in MYD88. In Aim 2, we will optimize validated hit compounds targeting MYD88 L265P. Data generated from this partnership will provide a solid scientific platform for therapeutic development targeting the oncogenic MYD88 L265P while sparing WT MYD88, which is critical for both innate and adaptive immunity. This work addresses the unmet clinical need to target MYD88 L265P directly and advances drug development against mutation-specific drivers.