Search

Building performance models for software services in DevOps is costly and error-prone. Accurate service demand distribution estimation is critical to precisely modeling queueing behaviors and performance prediction. However, current estimation methods focus on capturing the mean service demand, disregarding higher-order moments of the distribution that still can largely affect prediction accuracy. To address this limitation, we propose to estimate higher moments of the service demand distribution for a microservice from monitoring traces. We first generate a closed queueing model to abstract software performance and use it to model the departure process of requests completed by the software service as a Markovian arrival process (MAP). This allows formulating the estimation of service demand into an optimization problem, which aims to find the first multiple moments of the service demand distribution that maximize the likelihood of the MAP using generated the measured inter-departure times. We then estimate the service demand distribution for different classes of service with a maximum likelihood algorithm and novel heuristics to mitigate the computational cost of the optimization process for scalability. We apply our method to real traces from a microservice-based application and demonstrate that its estimations lead to greater prediction accuracy than exponential distributions assumed in traditional service demand estimation approaches for software services.

Introduction: Chronic myeloid leukemia (CML) is a kind of malignant tumor formed by the clonal proliferation of bone marrow hematopoietic stem cells. BCR-ABL fusion protein, found in more than 90% of patients, is a vital target for discovering anti-CML drugs. Up to date, imatinib is the first BCR-ABL tyrosine kinase inhibitor (TKI) approved by the FDA for treating CML. However, the drug resistance problems appeared for many reasons, especially the T135I mutation, a "gatekeeper" of BCR-ABL. Currently, there is no long-term effective and low side effect drug in clinical. Methods: This study intends to find novel TKIs targeting BCR-ABL with high inhibitory activity against T315I mutant protein by combining artificial intelligence technology and cell growth curve, cytotoxicity, flow cytometry and western blot experiments. Results: The obtained compound was found to kill leukemia cells, which had good inhibitory efficacy in BaF3/T315I cells. Compound No 4 could induce cell cycle arrest, cause autophagy and apoptosis, and inhibit the phosphorylation of BCR-ABL tyrosine kinase, STAT5 and Crkl proteins. Conclusion: The results indicated that the screened compound could be used as a lead compound for further research to discover ideal chronic myeloid leukemia therapeutic drugs.

To ensure the quality of frequent releases in DevOps context, performance models enable system performance simulation and prediction. However, building performance models for microservice or serverless-based applications in DevOps is costly and error-prone. Thus, we propose to employ model discovery learning for performance models automatically. To generate basic models to represent the application, we first introduce performance-related TOSCA models as architectural models. Then we transform TOSCA models into layered queueing network models. A main challenge of performance model generation is model parametrization. We propose to learn parametric dependencies from monitoring data and systems analysis to capture the relationship between input data and resource demand. With frequent releases of new features, we consider employing detecting parametric dependencies incrementally to keep updating performance models in each iteration.

Building performance models for microservices applications in DevOps is costly and error-prone. Accurate service demand distribution estimation is critical to performance model parameterization. However, traditional service demand estimation methods focus on capturing the mean service demand, disregarding higher-order moments of the distribution. To address this limitation, we propose to estimate higher moments of the service demand distribution for a microservice from monitoring traces. We first generate a closed queueing model to abstract a microservice and model the departure process at the queue node as a Markovian arrival process. This allows formulating the estimation of service demand as an optimization problem, which aims to find the optimal parameters of the first multiple moments of the service demand distribution based on the inter-departure times. We then estimate the service demand distribution with a novel maximum likelihood algorithm, and heuristics to mitigate the computational cost of the optimization process for scalability. We apply our method to real traces from a microservice-based application and demonstrate that its estimations lead to greater prediction accuracy than exponential distributions assumed in traditional service demand estimation approaches.

Epigenetic abnormalities contribute significantly to the development and progression of gastric cancer. However, the underlying regulatory networks from oncogenic signaling pathway to epigenetic dysregulation remain largely unclear. Here we showed that STAT3 signaling, one of the critical links between inflammation and cancer, acted as a control pathway in gastric carcinogenesis. STAT3 aberrantly transactivates the epigenetic kinase mitogen- and stress-activated protein kinase 1 (MSK1), thereby phosphorylating histone H3 serine10 (H3S10) and STAT3 itself during carcinogen-induced gastric tumorigenesis. We further identified the calcium pathway transcription factor NFATc2 as a novel downstream target of the STAT3-MSK1 positive-regulating loop. STAT3 forms a functional complex with MSK1 at the promoter of NFATc2 to promote its transcription in a H3S10 phosphorylation-dependent way, thus affecting NFATc2-related inflammatory pathways in gastric carcinogenesis. Inhibiting the STAT3/MSK1/NFATc2 signaling axis significantly suppressed gastric cancer cell proliferation and xenograft tumor growth, which provides a potential novel approach for gastric carcinogenesis intervention by regulating aberrant epigenetic and transcriptional mechanisms.

Autonomous cars employed as mobile base stations could provide communication networks in network-congested areas. In this paper, the authors leverage emerging deep reinforcement learning (DRL) techniques for enabling autonomous cars control and present a novel and highly effective DRL-based control framework called DRL-C3.

Experiment was performed to investigate the blast resistant performance of composite blast resistant wall subjected to variable blast loading. Then a numerical method was used to analyze the deformation process. In addition, an optimal design has been conducted to reduce the permanent maximum displacement and improve the stability of energy absorption. Effect of several design parameters, i.e. face-sheet thickness, concrete thickness, frame spacing are discussed. The research results show that composite blast resistant wall tend to have localized center dishing on the surface subjected to lower level blast loading. With the loading increasing, the center dishing expands to boundary. The tensile deformation at the center transform to the shear deformation at the boundary. When the blast loading is increasing continually, the penetration and the global failure occurs. After optimization of the structure parameters, the performance of the composite blast resistant wall is improved, and the optimization results can be used to guide the practical design.

The software reliability model is the basis for the quantitative analysis and prediction of software reliability. In recent years, neural networks due to its generalization and learning ability have been widely applied in the field of software reliability modeling. However, the slow convergence and local minimum of neural networks may cause unsuccessful prediction. Therefore, this paper presents a software reliability combination model based on genetic optimization BP neural network. This model uses three classical software reliability models as the input of BP neural network, and then uses the genetic algorithm optimization to automatically configure and optimize the weight and the thresholds. The results of experiments show that the model proposed has better fitting effect and prediction ability than other similar models.

The improper use of pointers in C/C++ programming language brings about a lot of memory-related issues. In this paper, causes of four kinds of memory defects are analyzed and summarized. Besides, a novel triad memory model has been proposed. Based on the model and the variable life cycle methodology, an approach for inner-procedure and inter-procedure detection has been presented too. Eventually, the prototype CAnalyzer is implemented on the basis of Clang static analyzer. Experiment results show that CAnalyzer can effectively detect the four types of memory defects.

In China, landscape had been altered from agricultural land to forest over an area of 219.37 million hectares during the last 50 years. Changes in land use can have significant impacts on soil characteristics, microbial properties and the transformation of soil organic matter, total nitrogen and associated nutrients, including phosphorus (P). P is integrant during the large number elements for plants. P the plants needed is mainly obtained from the soil P pool. In order to assess the impacts of Larch (Larix Olgensis) afforestation on soil P, the authors used Dunhua City in Jilin Province as a case region to investigate the soil P during different afforestation periods(l, 3, 8, 12, 18, 20, 33 years old), with an adjacent 250-year-old natural forest for the possible soil P storing in northeast China. We mainly analyzed the variation of soil P dynamics in different ages. Results showed that total P decreased with stand ages. The content of total P was closed with natural forest in the first years then decreased to a minimum level of 0.4378 g/kg in 12 years. In 18 years, the soil total P reached to a level as high as that in the former cropland, i.e. 0.811 g/kg then gradually depleted. Available P accumulated after afforestation. Available P concentration reached the maximum value of 153.96 mg/kg in 12 years. Within the short time span of 33 years, the contents of available P was closed to natural forest. The proportion of available P occupied in total P increased with age classes. The ratio was 0.24%-84.33%. Based on this, we made correlation analysis between available P and total P. It showed that there was no significant relation between them. Finally, factors that influenced the availability of P were briefly commented.

Diamond-Blackfan anaemia is a congenital bone marrow failure syndrome that is characterized by red blood cell aplasia. The disease has been associated with mutations or large deletions in 11 ribosomal protein genes including RPS7, RPS10, RPS17, RPS19, RPS24, RPS26, RPS29, RPL5, RPL11, RPL26 and RPL35A as well as GATA1 in more than 50% of patients. However, the molecular aetiology of many Diamond-Blackfan anaemia cases remains to be uncovered. To identify new mutations responsible for Diamond-Blackfan anaemia, we performed whole-exome sequencing analysis of 48 patients with no documented mutations/deletions involving known Diamond-Blackfan anaemia genes except for RPS7, RPL26, RPS29 and GATA1. Here, we identified a de novo splicing error mutation in RPL27 and frameshift deletion in RPS27 in sporadic patients with Diamond-Blackfan anaemia. In vitro knockdown of gene expression disturbed pre-ribosomal RNA processing. Zebrafish models of rpl27 and rps27 mutations showed impairments of erythrocyte production and tail and/or brain development. Additional novel mutations were found in eight patients, including RPL3L, RPL6, RPL7L1T, RPL8, RPL13, RPL14, RPL18A and RPL31. In conclusion, we identified novel germline mutations of two ribosomal protein genes responsible for Diamond-Blackfan anaemia, further confirming the concept that mutations in ribosomal protein genes lead to Diamond-Blackfan anaemia.

Transient abnormal myelopoiesis (TAM) is a clonal preleukemic disorder that progresses to myeloid leukemia of Down syndrome (ML-DS) through the accumulation of genetic alterations. To investigate the mechanism of leukemogenesis in this disorder, a xenograft model of TAM was established using NOD/Shi-scid, interleukin (IL)-2Rγ(null) mice. Serial engraftment after transplantation of cells from a TAM patient who developed ML-DS a year later demonstrated their self-renewal capacity. A GATA1 mutation and no copy number alterations (CNAs) were detected in the primary patient sample by conventional genomic sequencing and CNA profiling. However, in serial transplantations, engrafted TAM-derived cells showed the emergence of divergent subclones with another GATA1 mutation and various CNAs, including a 16q deletion and 1q gain, which are clinically associated with ML-DS. Detailed genomic analysis identified minor subclones with a 16q deletion or this distinct GATA1 mutation in the primary patient sample. These results suggest that genetically heterogeneous subclones with varying leukemia-initiating potential already exist in the neonatal TAM phase, and ML-DS may develop from a pool of such minor clones through clonal selection. Our xenograft model of TAM may provide unique insight into the evolutionary process of leukemia.

Diamond-Blackfan anemia (DBA) is an inherited bone marrow disease. The condition is characterized by anemia that usually presents during infancy or early childhood and congenital malformation. Several reports show that DBA is associated with mutations in the ribosomal protein (RP) genes, RPS19, RPS24, RPS17, RPL35A, RPL5, RPL11, and RPS7. Recently, 5 and 12 patients with mutations in RPS10 and RPS26, respectively, were identified in a cohort of 117 DBA probands. Therefore, we screened the DBA patients who were negative for mutations in these DBA genes for mutations in RPS10 and RPS26. The present case report describes the identification of the first Japanese DBA patient with a novel mutation in RPS10.

Abstract 535 Background Transient abnormal myelopoiesis (TAM) represents a self-limited proliferation exclusively affecting perinatal infants with Down syndrome (DS), morphologically and immunologically characterized by immature blasts indistinguishable from acute megakaryoblastic leukemia (AMKL). Although spontaneous regression is as a rule in most cases, about 20–30% of the survived infants develop non-self-limited AMKL (DS-AMKL) 3 to 4 years after the remission. As for the molecular pathogenesis of these DS-related myeloid proliferations, it has been well established that GATA1 mutations are detected in virtually all TAM cases as well as DS-AMKL. However, it is still open to question whether a GATA1 mutation is sufficient for the development of TAM, what is the cellular origin of the subsequent AMKL, whether additional gene mutations are required for the progression to AMKL, and if so, what are their gene targets, although several genes have been reported to be mutated in occasional cases with AMKL, including JAK2/3, TP53 and FLT3. Methods To answer these questions, we identify a comprehensive spectrum of gene mutations in TAM/AMKL cases using whole genome sequencing of three trio samples sequentially obtained at initial presentation of TAM, during remission and at the subsequent relapse phase of AMKL. Whole exome sequencing was also performed for TAM (N=16) and AMKL (N=15) samples, using SureSelect (Agilent) enrichment of 50M exomes followed by high-throughput sequencing. The recurrent mutations in the discovery cohort were further screened in an extended cohort of DS-AMKL (N = 35) as well as TAM (N = 26) and other AMKL cases (N = 19) using target deep sequencing. Results TAM samples had significantly fewer numbers of somatic mutations compared to AMKL samples with the mean numbers of all mutations of 30 (1.0/Mb) and 180 (6.0/Mb) per samples in whole genome sequencing or non-silent somatic mutations of 1.73 and 5.71 per sample in whole exome sequencing in TAM and AMKL cases, respectively (p=0.001). Comprehensive detections of the full spectrum of mutations together with subsequent deep sequencing of the individual mutations allowed to reveal more complicated clonological pictures of clonal evolutions leading to AMKL. In every patient, the major AMKL clones did not represent the direct offspring from the dominant TAM clone. Instead, the direct ancestor of the AMKL clones could be back-traced to a more upstream branch-point of the evolution before the major TAM clone had appeared or, as previously reported, to an earlier founder having an independent GATA1 mutation. Intratumoral heterogeneity was evident at the time of diagnosis as the presence of major subpopulations in both TAM and AMKL populations, which were more often than not characterized by RAS pathway mutations. While GATA1 was the only recurrent mutational target in the TAM phase, 8 genes were recurrently mutated in AMKL samples in whole genome/exome sequencing, including NRAS, TP53 and other novel gene targets that had not been previously reported to be mutated in other neoplasms. The recurrent mutations found in the discovery cohort, in addition to known mutational targets in myeloid malignancies, were screened in an extended cohort of DS-associated myeloid disorders (N=61) as well as other AMKL cases, using high-throughput sequencing of SureSelect-captured and/or PCR amplified targets. Secondary mutations other than GATA1 mutations were found in 3 out of 26 TAM, 20 out of 35 DS-AMKL and 4 out of 19 other AMKL cases. Conclusion TAM is characterized by a paucity of somatic mutations and thought to be virtually caused by a GATA1 mutation in combination with constitutive trisomy 21. Subsequent AMKL evolved from a minor independent subclone acquiring additional mutations. Secondary genetic hits other than GATA1 mutations were common, where deregulated epigenetic controls as well as abnormal signaling pathway mutations play a major role. Disclosures: No relevant conflicts of interest to declare.

Abstract 984 Diamond-Blackfan anemia (DBA) is a congenital bone marrow failure syndrome, characterized by red blood cell aplasia, macrocytic anemia, and increased risk of malignancy. Approximately 90% of patients present during the first year of life or in early childhood. About 40–50% of DBA cases are familial with autosomal dominant, while the remainder is sporadic cases whose mode of inheritance is largely unknown. Although anemia is the most prominent feature of DBA, up to 40% of patients also accompany other symptoms including growth retardation and/or a variety of congenital malformations. Recent studies have shown that the disease could be associated with heterozygous mutations in ribosomal protein (RP) genes, including six small subunit RP genes RPS7, RPS10, RPS17, RPS19, RPS24, and RPS26 as well as four large subunit RP genes RPL5, RPL11, RPL26, and RPL35A, which collectively account for about 50% of patients with DBA. In addition, germline mutations in the GATA1 gene encoding a hematopoietic transcription factor, have been also reported in two DBA families. However, it is clear that the molecular etiology of many DBA cases remains to be covered. To identify new mutations that are responsible for DBA, we performed whole-exome sequencing on 40 DBA patients with no documented mutations/deletions involving known DBA genes. After excluding all variants registered in the 1000 Genomes Project, or dbSNP131, or found in our inhouse SNP database, we searched for non-synonymous mutations involving RP genes as possible candidate for novel DBA genes. In this study, we identified probable pathogenic mutations in two novel RP genes, RPS27 and RPL27 in two patients. The first case was a 1-year-old girl who harbored a single nucleotide substitution at the splice acceptor site in intron 1 of RPL27 (c.-2–1G>A), which results in splicing error. She had atrial septal defect and pulmonary stenosis, and responded to steroid treatment. The second case was a 2-year-old girl carrying a frameshift deletion of RPS27 (c.90delC, p.Tyr31ThrfsX5), leading to a premature truncation. This patient had no abnormalities and responded to steroid treatment. An additional five missense SNVs affecting single cases was identified in five genes, including RPL3L, RPL8, RPL13, RPL18A, and RPL31, together with two in-frame deletions of RPL6 and RPL14 in two patients, which cause deletion of a single amino-acid. However, the pathological significance in these 7 cases is uncertain. In the remaining 31 patients, no mutations were detected in RP genes. In conclusion, we identified novel germline mutations of RP genes that could be responsible for DBA, further confirming the concept that the RP genes are common targets of germline mutations in DBA patients and also suggested the presence of non-RP gene targets for DNA. Disclosures: No relevant conflicts of interest to declare.

Abstract 1491 Five to ten percent of neonates with Down syndrome (DS) develop transient abnormal myelopoiesis (TAM), which is characterized by a rapid growth of abnormal blast cells with the erythroid/megakaryocytic phenotype. In most cases, it resolves spontaneously within three months. However, 20 to 30% of patients develop acute megakaryocytic leukemia referred to as myeloid leukemia of DS (ML-DS) within four years. Blast cells in most of TAM and ML-DS patients have mutations in exon 2 of the gene encoding transcription factor GATA1, which is essential for the normal development of erythroid and megakaryocytic cells. These mutations lead to expression of a truncated GATA1 protein lacking the N-terminal 83 amino acids (GATA1s). However, the molecular mechanism whereby GATA1s contributes to the genesis of TAM and ML-DS remains elusive. From 2003 to 2010, we screened GATA1 mutations in clinical samples obtained from 106 patients with TAM upon request from referring hospitals. We performed direct sequencing analysis using cDNA prepared form total RNA extracted from white blood cells in peripheral blood and genomic DNA. Acquired GATA1 mutations were detected in a total of 99 (93.4%) patients among them. Most of these mutations were coding GATA1s. However, the mutations coding GATA1 mutants with internal deletion of 43 and 15 amino acids were detected in five and one patients, and we referred to them as GATA-ID type-1 and GATA1-ID type-2, respectively. Sequences of exon 3 were discrepant with that expected from the corresponding cDNA sequence. Instead, a 2-nucleotide insertion, a 2-nucleotide deletion and a 7-nucleotide deletion with a single-nucleotide substitution were detected in exon 3 of the GATA1 gene in the GATA1-ID type-1 patients, and a 21-nucleotide deletion encompassing the boundary of intron 2 and exon 3 was found in the GATA-ID type-2 patient. These findings suggest that the mutations in exon 3 lead to expression of alternative splicing forms of mRNA encoding GATA1-ID. To investigate the role of GATA1-ID on the pathogenesis of TAM and ML-DS, we next expressed GATA1-ID by retrovirus in a ML-DS cell line, KPAM1, or primary GATA1-deficient fetal megakaryocytic progenitors. Expectedly, the cell growth was markedly reduced upon transduction with full-length GATA1-expression retrovirus. However, ectopic expression of GATA1-ID type-1 and type-2 failed to restrict the proliferation of cells as similar as expression of GATAs. Interestingly, all these patients had high white blood cell counts in the peripheral blood at diagnosis. Furthermore, three patients developed liver fibrosis and two had effusion. All these factors are significantly associated with early death. Three patients were treated with low dose of Ara-C. However, early death occurred in three patients. In this study we found that GATA1-ID proteins are created by somatic mutations in DS patients and contribute to the generation of TAM phenotypes. Our newly identified GATA1-ID mutants have highlighted a much narrower set of sequences responsible for the pathogenesis of than was previously suggested. The missing region identified by the GATA1-ID proteins contains a consensus motif (LxCxE, amino acids 81–85) essential for the interaction with pRB protein and this motif is also lost in GATA1s. Our results suggested the lack of Rb-GATA1 interaction as the most likely pathogenesis for onset of TAM. Disclosures: No relevant conflicts of interest to declare.

Abstract 4231 Introduction: Fifty percent of Diamond–Blackfan anemia (DBA) patients possess mutations in ribosomal protein genes. Although several ribosomal protein genes, RPL5, RPL11, RPL35A, RPS7, RPS10, RPS17, RPS19, RPS24, and RPS26, have been reported to be mutated in some DBA patients, including point mutations, nonsense mutations, deletions, splice site mutations, and translocations, other DBA patients appear to have intact ribosomal protein genes. To identify new mutations in ribosomal protein genes from a different aspect, we focused on extensive deletions in these genes, such as mutations involving loss of a whole allele. In this study, we applied quantitative genomic PCR, and successfully developed a convenient method for detecting extensive deletions designated the “DBA gene copy number assay”. Methods: DBA patients should have an intact allele and a mutated allele for the responsible ribosomal protein gene, meaning that they will have an abnormal karyotype (gene copy number of N) if they have an extensive deletion. We attempted to clarify the copy numbers of ribosomal protein genes by the difference in a 1-cycle delay of threshold in a quantitative PCR (q-PCR) assay. To detect extensive deletions, at least 2 sets of gene-specific primers for each DBA responsible gene (RPL5, RPL11, RPL35A, RPS7, RPS10, RPS17, RPS19, RPS24, and RPS26) were prepared. Appropriate primers to fit the setting that the threshold cycle (Ct) of the q-PCR should occur within 1 cycle of the Ct scores of other primer sets were selected. After validation, we identified 6, 3, 4, 3, 3, 6, 9, 3, and 2 specific primer sets for RPL5, RPL11, RPL35A, RPS7, RPS10, RPS17, RPS19, RPS24, and RPS26, respectively. By simply looking at the q-PCR amplification curves by eye, we were easily able to judge the copy numbers of 2N (normal) or N (abnormal) for the ribosomal protein genes. Results: We performed the DBA gene copy number assay for 14 randomly selected undiagnosed patients from the Japanese DBA genomic resource at the University of Hirosaki, who had no mutations by genomic sequencing analyses. For each case, all the DBA responsible genes were confirmed using the diagnostic primers. The results of the DBA gene copy number assays revealed that 5 of the 14 probands (36%) had an extensive deletion in one of the DBA responsible genes. As an interesting case among the 5 positive cases, we confirmed an extensive deletion in the RPS19 gene. The Ct scores for 4 of the 9 primer sets for RPS19 demonstrated a 1-cycle delay, while the scores for the other 5 primer sets were normal. By genomic PCR amplification analyses, we identified a deletion from nt. -1400 to +5757 (7157 nucleotides) in the RPS19 gene. The deleted region included the promoter region, and exons 1, 2, and 3 of the RPS19 gene. The remaining 4 cases were 1 proband with an RPL5 deletion, 1 with an RPL35A deletion and 2 with RPS17 deletions. In particular, the extensive deletions in the RPL5 and RPS17 alleles are the first such cases reported. Discussion: Since it has been difficult to address the loss of a whole allele in DBA, such mutations have not been precisely examined within the DBA responsible genes. Our data suggest that extensive deletions in ribosomal protein genes comprise a significant proportion of DBA cases in Japan. Our novel method could become a useful tool for screening the gene copy numbers of ribosomal protein genes, and for identifying new pathological mutations. Disclosures: No relevant conflicts of interest to declare.

DevOps is an emerging paradigm that reduces the barriers between developers and operations teams to offer continuous fast delivery and enable quick responses to changing requirements within the software life cycle. A significant volume of activity has been carried out in recent years with the aim of coupling DevOps stages with tools and methods to improve the quality of the produced software and the underpinning delivery methodology. While the research community has produced a sustained effort by conducting numerous studies and innovative development tools to support quality analyses within DevOps, there is still a limited cohesion between the research themes in this domain and a shortage of surveys that holistically examine quality engineering work within DevOps. In this paper, we address the gap by comprehensively surveying existing efforts in this area, categorizing them according to the stage of the DevOps lifecycle to which they primarily contribute. The survey holistically spans across all the DevOps stages, identify research efforts to improve architectural design, modeling and infrastructure-as-code, continuous-integration/continuous-delivery (CI/CD), testing and verification, and runtime management. Our analysis also outlines possible directions for future work in quality-aware DevOps, looking in particular at AI for DevOps and DevOps for AI software.