Search

Bruton's tyrosine kinase proteolysis-targeting chimeras (BTK-PROTACs) have emerged as a promising approach to address the limitations of BTK inhibitors. However, conducting the rational discovery of orally bioavailable BTK-PROTACs presents significant challenges. In this study, dimensionality reduction analysis and model molecule validation were utilized to identify some key structural features for improving the oral absorption of BTK-PROTACs. The results were applied to optimize the newly discovered BTK-PROTACs

CXC chemokine receptors 1 (CXCR1) and 2 (CXCR2) have been demonstrated to have critical roles in cancer metastasis. Because they share high homology sequences, it is still unclear how to design selective CXCR1 or CXCR2 antagonists. Based on a pharmacophore model we built, compound 2 bearing a 1,5-dihydro-4H-imidazol-4-one scaffold was identified as a selective CXCR2 antagonist with a low CXCR1 antagonism preference. Further optimization and structure-activity relationship studies led to compound C5 that overcame the disadvantages of compound 2 and performed with higher selectivity. It showed excellent oral bioavailability and in vitro anticancer metastasis activity. Further dynamic simulation of the molecular protein complex showed that the amino acid residue K320 of CXCR2 contributed most to the selectivity of C5. This study provides important clues for the design of new CXCR2 selective antagonists, and C5 can be a molecular tool for investigating the difference in the biological function of CXCR1 and CXCR2.

Currently, deep learning plays an indispensable role in many fields, including computer vision, natural language processing, and speech recognition. Convolutional Neural Networks (CNNs) have demonstrated excellent performance in computer vision tasks thanks to their powerful feature extraction capability. However, as the larger models have shown higher accuracy, recent developments have led to state-of-the-art CNN models with increasing resource consumption. This paper investigates a conceptual approach to reduce the memory consumption of CNN inference. Our method consists of processing the input image in a sequence of carefully designed tiles within the lower subnetwork of the CNN, so as to minimize its peak memory consumption, while keeping the end-to-end computation unchanged. This method introduces a trade-off between memory consumption and computations, which is particularly suitable for high-resolution inputs. Our experimental results show that MobileNetV2 memory consumption can be reduced by up to 5.3 times with our proposed method. For ResNet50, one of the most commonly used CNN models in computer vision tasks, memory can be optimized by up to 2.3 times.

Non-small-cell lung cancer (NSCLC), a subtype of lung cancer, is one of the leading causes of cancer death in both men and women around the world. Circulating tumour DNA (ctDNA), tyrosine kinase inhibitors (TKIs), and immunotherapy have transformed our understanding of NSCLC and its treatment, from diagnosis to targeted NSCLC therapeutics. Quantifying ctDNA is convenient and precise, making clinical decisions easier. TKI-based targeted therapy and immunotherapy have also enhanced the quality of life of NSCLC patients. This article gives an update on ctDNA technologies and their implications for therapeutic options, including medications targeting the epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) with TKIs such as osimertinib and lorlatinib, the emergence of various resistance mechanisms, the control of programmed cell death-1 (PD- 1), programmed cell death ligand-1 (PD-L1), and cytotoxic T-lymphocyte antigen-4 (CTLA-4) by immune checkpoint inhibitors (ICIs) in immunotherapy and the blood tumor mutational burden (bTMB) calculated by ctDNA assay as a novel biomarker for immunotherapy. NSCLC patients, on the other hand, nevertheless confront numerous hurdles. To produce more effective medications or therapies to treat NSCLC, additional research and trials are needed.

Oncogenic KIT or PDGFRA tyrosine kinase mutations are compelling therapeutic targets in most gastrointestinal stromal tumors (GISTs), and the KIT inhibitor, imatinib, is therefore standard of care for patients with metastatic GIST. However, some GISTs lose expression of KIT oncoproteins, and therefore become KIT-independent and are consequently resistant to KIT-inhibitor drugs. We identified distinctive biologic features in KIT-independent, imatinib-resistant GISTs as a step towards identifying drug targets in these poorly understood tumors. We developed isogenic GIST lines in which the parental forms were KIT oncoprotein-dependent, whereas sublines had loss of KIT oncoprotein expression, accompanied by markedly downregulated expression of the GIST biomarker, protein kinase C-theta (PRKCQ). Biologic mechanisms unique to KIT-independent GISTs were identified by transcriptome sequencing, qRT-PCR, immunoblotting, protein interaction studies, knockdown and expression assays, and dual-luciferase assays. Transcriptome sequencing showed that cyclin D1 expression was extremely low in two of three parental KIT-dependent GIST lines, whereas cyclin D1 expression was high in each of the KIT-independent GIST sublines. Cyclin D1 inhibition in KIT-independent GISTs had anti-proliferative and pro-apoptotic effects, associated with Rb activation and p27 upregulation. PRKCQ, but not KIT, was a negative regulator of cyclin D1 expression, whereas JUN and Hippo pathway effectors YAP and TAZ were positive regulators of cyclin D1 expression. PRKCQ, JUN, and the Hippo pathway coordinately regulate GIST cyclin D1 expression. These findings highlight the roles of PRKCQ, JUN, Hippo, and cyclin D1 as oncogenic mediators in GISTs that have converted, during TKI-therapy, to a KIT-independent state. Inhibitors of these pathways could be effective therapeutically for these now untreatable tumors.

In recent years, Convolutional Neural Networks (CNNs) have enabled unprecedented progress on a wide range of computer vision tasks. However, training large CNNs is a resource-intensive task that requires specialized Graphical Processing Units (GPU) and highly optimized implementations to get optimal performance from the hardware. GPU memory is a major bottleneck of the CNN training procedure, limiting the size of both inputs and model architectures. In this paper, we propose to alleviate this memory bottleneck by leveraging an under-utilized resource of modern systems: the device to host bandwidth. Our method, termed CPU offloading, works by transferring hidden activations to the CPU upon computation, in order to free GPU memory for upstream layer computations during the forward pass. These activations are then transferred back to the GPU as needed by the gradient computations of the backward pass. The key challenge to our method is to efficiently overlap data transfers and computations in order to minimize wall time overheads induced by the additional data transfers. On a typical work station with a Nvidia Titan X GPU, we show that our method compares favorably to gradient checkpointing as we are able to reduce the memory consumption of training a VGG19 model by 35% with a minimal additional wall time overhead of 21%. Further experiments detail the impact of the different optimization tricks we propose. Our method is orthogonal to other techniques for memory reduction such as quantization and sparsification so that they can easily be combined for further optimizations.

[Image: see text] CXC chemokine receptors 1 (CXCR1) and 2 (CXCR2) have been demonstrated to have critical roles in cancer metastasis. Because they share high homology sequences, it is still unclear how to design selective CXCR1 or CXCR2 antagonists. Based on a pharmacophore model we built, compound 2 bearing a 1,5-dihydro-4H-imidazol-4-one scaffold was identified as a selective CXCR2 antagonist with a low CXCR1 antagonism preference. Further optimization and structure–activity relationship studies led to compound C5 that overcame the disadvantages of compound 2 and performed with higher selectivity. It showed excellent oral bioavailability and in vitro anticancer metastasis activity. Further dynamic simulation of the molecular protein complex showed that the amino acid residue K320 of CXCR2 contributed most to the selectivity of C5. This study provides important clues for the design of new CXCR2 selective antagonists, and C5 can be a molecular tool for investigating the difference in the biological function of CXCR1 and CXCR2.

Multiple myeloma (MM) is a highly malignant hematologic cancer that occurs when an atypical plasma cell develops in the bone marrow and reproduces quickly. Despite varies of new drugs have been developed or under clinic trial, MM is still essentially incurable, while XPO1 inhibition has emerged as a promising therapeutic strategy in the treatment of MM. Using the second-generation XPO1 inhibitor KPT-8602 as the lead compound, structure-based optimization provided D4 with high anti-proliferation efficacy (IC

In this paper, we propose a method to reduce the memory requirement of Convolutional Neural Networks (CNN) in the inference phase. Before feeding an input image into the CNN model, input image will split evenly to several sub-images and feed them into models respectively and combine the output after a certain layer.

Non-small-cell lung cancer (NSCLC), a main subtype of lung cancer, is one of the most common causes of cancer death in men and women worldwide. Circulating tumor DNA (ctDNA), tyrosine kinase inhibitors (TKIs) and immunotherapy have revolutionized both our understanding of NSCLC, from its diagnosis to targeted NSCLC therapies, and its treatment. ctDNA quantification confers convenience and precision to clinical decision making. Furthermore, the implementation of TKI-based targeted therapy and immunotherapy has significantly improved NSCLC patient quality of life. This review provides an update on the methods of ctDNA detection and its impact on therapeutic strategies; therapies that target epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) using TKIs such as osimertinib and lorlatinib; the rise of various resistant mechanisms; and the control of programmed cell death-1 (PD-1), programmed cell death ligand-1 (PD-L1), and cytotoxic T-lymphocyte antigen-4 (CTLA-4) by immune checkpoint inhibitors (ICIs) in immunotherapy; blood tumor mutational burden (bTMB) calculated by ctDNA assay as a novel biomarker for immunotherapy. However, NSCLC patients still face many challenges. Further studies and trials are needed to develop more effective drugs or therapies to treat NSCLC.

Programmed cell death-1/programmed cell death ligand 1 (PD-1/PD-L1) is one of the most promising targets in the field of immune checkpoint blockade therapy. Beginning with our exploration of linkers and structure-activity relationship research, we found that the aromatic ring could replace the linker and aryl group to maintain the satisfactory activity of classic triaryl scaffold inhibitor. Based on previous studies, we designed and synthesized a series of C2-symmetric phenyl-linked compounds, and further tail optimization afforded the inhibitors, which displayed promising inhibitory activity against the PD-1/PD-L1 interaction with IC50 value at the single nanomolar range (C13-C15). Further cell-based PD-1/PD-L1 blockade bioassays indicated that these C2-symmetric molecules could significantly inhibit the PD-1/PD-L1 interaction at the cellular level and restore T cells' immune function at the safety concentrations. The discovery of these phenyl-linked symmetric small molecules showed the potential of simplified-linker and C2-symmetric strategy and provided a basis for developing symmetric small molecule inhibitors of PD-1/PD-L1 interaction. Moreover, C13 and C15 performed stable binding modes to PD-L1 dimeric after computational docking and dynamic simulation, which may serve as a good starting point for further development.

Convolutional Neural Networks (CNN) have demonstrated state-of-the-art results on various computer vision problems. However, training CNNs require specialized GPU with large memory. GPU memory has been a major bottleneck of the CNN training procedure, limiting the size of both inputs and model architectures. Given the ubiquity of CNN in computer vision, optimizing the memory consumption of CNN training would have wide spread practical benefits. Recently, reversible neural networks have been proposed to alleviate this memory bottleneck by recomputing hidden activations through inverse operations during the backward pass of the backpropagation algorithm. In this paper, we push this idea to extreme and design a reversible neural network with minimal training memory consumption. The result demonstrated that we can train CIFAR10 dataset on Nvidia GTX750 GPU only with 1GB memory and achieve 93% accuracy within 67 minutes.

This article has been withdrawn by the authors. Some of the SDHA enzyme activity data were flawed and were not performed and analyzed correctly. The withdrawing authors are in the process of correcting the data and re-evaluating them for resubmission.