Lung cancer (LC) is the leading cause of cancer death in the US, claiming over 160,000 lives annually. Although CT screening has been shown to be efficacious in reducing mortality, the limited access to screening programs among high-risk individuals and the high number of false positives contribute to low survival rates and increased healthcare costs. As a result, there is an urgent need for preventative therapeutics and novel interception biomarkers that would enhance current methods for detection of early-stage LC. This thesis addresses this challenge by examining the hypothesis that transcriptomic changes preceding the onset of LC can be identified by studying bronchial premalignant lesions (PMLs) and the normal-appearing airway epithelial cells altered in their presence (i.e., the PML-associated airway field of injury). PMLs are the presumed precursors of lung squamous cell carcinoma (SCC) whose presence indicates an increased risk of developing SCC and other subtypes of LC. Here, I leverage high-throughput mRNA and miRNA sequencing data from bronchial brushings and lesion biopsies to develop biomarkers of PML presence and progression, and to understand regulatory mechanisms driving early carcinogenesis. First, I utilized mRNA sequencing data from normal-appearing airway brushings to build a biomarker predictive of PML presence. After verifying the power of the 200-gene biomarker to detect the presence of PMLs, I evaluated its capacity to predict PML progression and detect presence of LC (Aim 1). Next, I identified likely regulatory mechanisms associated with PML severity and progression, by evaluating miRNA expression and gene coexpression modules containing their targets in bronchial lesion biopsies (Aim2). Lastly, I investigated the preservation of the PML-associated miRNAs and gene modules in the airway field of injury, highlighting an emergent link between the airway field and the PMLs (Aim 3). Overall, this thesis suggests a multi-faceted utility of PML-associated genomic signatures as markers for stratification of high-risk smokers in chemoprevention trials, markers for early detection of lung cancer, and novel chemopreventive targets, and yields valuable insights into early lung carcinogenesis by characterizing mRNA and miRNA expression alterations that contribute to premalignant disease progression towards LC. / 2020-01-25

Cellular Thermal Shift Assay (CETSA) is a method of identifying protein-drug interactions by monitoring changes in protein thermal stability. CETSA is traditionally performed by using Western Blotting to examine the thermal stability shifts of a single protein of interest. By combining CETSA with Mass Spectrometry the shifts in thermal stability can be examined for an entire proteome in a single experiment in a technique known as CETSA-MS or Thermal Proteome Profiling (TPP). This can be used to identify targets of a compound of interest in order to further understand the compounds mechanism of interest, potentially making CETSA a powerful tool for target discovery. Here we attempt to develop a protocol by which CETSA can be used as a drug target discovery tool. Our work has allowed us to create a protocol that can reliably identify soluble drug targets. Our results demonstrate the capacity of CETSA to screen multiple compounds as well as to perform more in depth dose response studies, and highlight how future improvements could be made to the protocol

The ease or difficulty in solving a problemstrongly depends on the way it is represented. For example, consider the task of multiplying the numbers 12 and 24. Now imagine multiplying XII and XXIV. Both tasks can be solved, but it is clearly more difficult to use the Roman numeral representations of twelve and twenty-four. Humans excel at finding appropriate representations for solving complex problems. This is not true for artificial systems, which have largely relied on humans to provide appropriate representations. The ability to autonomously construct useful representations and to efficiently exploit them is an important challenge for artificial intelligence. This dissertation builds on a recently introduced graph-based approach to learning representations for sequential decision-making problems modeled as Markov decision processes (MDPs). Representations, or basis functions, forMDPs are abstractions of the problem’s state space and are used to approximate value functions, which quantify the expected long-term utility obtained by following a policy. The graph-based approach generates basis functions capturing the structure of the environment. Handling large environments requires efficiently constructing and utilizing these functions. We address two issues with this approach: (1) scaling basis construction and value function approximation to large graphs/data sets, and (2) tailoring the approximation to a specific policy’s value function. We introduce two algorithms for computing basis functions from large graphs. Both algorithms work by decomposing the basis construction problem into smaller, more manageable subproblems. One method determines the subproblems by enforcing block structure, or groupings of states. The other method uses recursion to solve subproblems which are then used for approximating the original problem. Both algorithms result in a set of basis functions from which we employ basis selection algorithms. The selection algorithms represent the value function with as few basis functions as possible, thereby reducing the computational complexity of value function approximation and preventing overfitting. The use of basis selection algorithms not only addresses the scaling problem but also allows for tailoring the approximation to a specific policy. This results in a more accurate representation than obtained when using the same subset of basis functions irrespective of the policy being evaluated. To make effective use of the data, we develop a hybrid leastsquares algorithm for setting basis function coefficients. This algorithm is a parametric combination of two common least-squares methods used for MDPs. We provide a geometric and analytical interpretation of these methods and demonstrate the hybrid algorithm’s ability to discover improved policies. We also show how the algorithm can include graphbased regularization to help with sparse samples from stochastic environments. This work investigates all aspects of linear value function approximation: constructing a dictionary of basis functions, selecting a subset of basis functions from the dictionary, and setting the coefficients on the selected basis functions. We empirically evaluate each of these contributions in isolation and in one combined architecture.

No description available.

学位プログラム名: 京都大学大学院思修館 / 京都大学 / 新制・課程博士 / 博士(総合学術) / 甲第23345号 / 総総博第18号 / 新制||総総||3(附属図書館) / 京都大学大学院総合生存学館総合生存学専攻 / (主査)教授 山敷 庸亮, 教授 杉山 弘, 教授 積山 薫 / 学位規則第4条第1項該当 / Doctor of Philosophy / Kyoto University / DGAM

Gene therapy has emerged as a revolutionary approach for treating genetic disorders, holding great promise for improving patient outcomes. Among the various viral vectors used for delivery of therapeutic transgenes, Adeno-Associated Viruses (AAVs) have gained prominence due to their favorable characteristics including low immunogenicity, long-term gene expression, and the ability to target both dividing and non-dividing cells. However, AAV’s are associated with the high costs of production and challenges with production of a high-quality virus, limiting AAV’s utilization and widespread use. In this study, we aimed to develop a high-throughput screening assay targeting AAV production enhancers, thus addressing the manufacturing obstacles and advancing the affordability and accessibility of gene therapies. To help overcome the limitations and expenses associated with AAV manufacturing, an innovative high-throughput screening assay was developed with the intent to identify cell culture additives/conditions which maximize AAV production. We optimized various parameters, including the transgene, producer and reporter cell lines, harvest timings and methods, and transduction techniques. The optimized screening assay was employed to evaluate novel compounds across several timings of addition, for their ability to enhance AAV production. Notably, several compounds indicated transfection enhancing capabilities up to 3.4-fold and the developed assays final variability was below 14%. Additionally, compound combinations were assessed to uncover potential additive and synergistic effects that could further enhance AAV productivity. In conclusion, our study presents a significant advancement in targeting the manufacturing challenges associated with AAV. By utilizing an optimized high-throughput screening assay, researchers and manufacturers can identify compounds that enhance AAV production, paving the way for cost-effective and scalable manufacturing processes. Ultimately, this progress holds the potential to improve the affordability, accessibility, and impact of gene therapies for patients worldwide.

In 2020, cancer impacted an estimated 1.8 million people and result in over 600,000 deaths in the United States. Some cancer treatments options are limited due to drug resistance, requiring additional drug development to improve patient survival rates. It is necessary to continuously develop new therapeutic approaches and identify novel targets, as cancer is ever-growing and adapting. Experimental research strategies have limitations when exploring how to target certain protein classes, including membrane-embedded or protein-protein bound, due to the complexity of their environments. These two domains of research are experimentally challenging to explore, and in silico research practices provide insight that would otherwise take years to study. Computer-aided drug design (CADD) routes can support the areas of drug discovery that are considered difficult to explore with experimental techniques. In this work, we provide research practices that are easily adaptable and translatable to other difficult protein targets and interfaces. First, we identified the morphological impact of a single-site mutation in the G-protein coupled receptor (GPCR), OR2T7, which had been identified as a novel prognostic marker for glioblastoma. Next, we explored the blockbuster target, Programmed Cell Death Protein 1 – (PD-1) and the agonistic vs antagonistic response that can be exploited for Non-Small Cell Lung Cancer (NSCLC) therapeutic development. Last, we explored the sphingolipid transport protein, Spns2, which has been demonstrated to be important in regulating the metastatic cancer enabling microenvironment. This work utilized molecular dynamics simulations (MDS) to explore the protein structure-function relationship for each protein of interest, allowing for the exploration of biophysical properties and protein dynamics. We identified that the D125V mutation in OR2T7 likely influences activation of the MAPK pathway by impacting G-protein binding via reducing the helical plasticity in the TM6 and TM7 regions. PD-1 was identified to have a domain near the PD-L1 binding interface that increases β-sheet stability and increases residue-residue distances with the membrane-proximal region within PD-1, thus leading to an active conformation. Lastly, Spns2 was identified to follow a rocker-switch transport model and provided preliminary insight into sphingolipid-Spns2 channel binding, interacting with residues Thr216, Arg227, and Met230, as well as highlighting the role of Arg119 in a salt-bridge network of interactions essential in substrate translocation. Collectively, this work illustrates the advantages of computational workflows in the drug discovery process and provides a framework that can be applied for additional GCPRs, transport proteins, or protein-protein interfaces to enhance and accelerate the CADD research. / Doctor of Philosophy / Cancer is an ever-evolving disease that requires continuous development of new treatment options. Experimental research strategies can be timely, expensive, or lack atomistic insight into drug development processes. Computer-aided drug design (CADD) routes provide research strategies to support areas of drug discovery that can be difficult to explore with experimental techniques. Membrane-bound proteins and protein-protein interfaces are two domains of research that are typically difficult to explore, and computational research practices provide insight that would otherwise take years to study. In this work, we provide research practices that are easily adaptable and translatable to other difficult protein targets and interfaces. First, we identified the impact of a single-site mutation in the G-protein coupled receptor (GPCR), OR2T7, which had been identified as a novel prognostic marker for glioblastoma. Next, we explored the blockbuster target, Programmed Cell Death Protein 1 – (PD-1) and active vs inactive states that can be exploited for Non-Small Cell Lung Cancer (NSCLC) therapeutic development. Last, we explored the sphingolipid transport protein, Spns2, which has been demonstrated to be important in metastatic cancer growth. This work utilized molecular dynamics simulations (MDS) to explore the protein structure-function relationship for each protein of interest, allowing for the exploration of biophysical properties and protein movement. We identified that the D125V mutation in OR2T7 likely influences activation of the MAPK pathway, which supports multiple cancer-regulation pathways, by impacting G-protein binding via reducing the structural flexibility. PD-1 was identified to have a domain near the PD-L1 binding interface that increases structural stability, thus leading to an upregulation of cancer survival pathways. Lastly, Spns2 analysis provided insight into movement involved in sphingolipid transport, provided preliminary insight into sphingolipid-Spns2 binding, as well as highlighting the role of Arg119 in a network of interactions essential in substrate translocation. Collectively, this work highlights the usefulness of computational workflows in the drug discovery process and provides a framework that can be utilized for additional GPCRs, transport proteins, or protein-protein interfaces to enhance and accelerate the CADD research.

Doctor of Philosophy / Startups can and do fail. For an entrepreneur, product developer, or researcher with a physical and capital-intensive product idea, this dissertation can serve as a resource to bridge the gaps between business, engineering, and design and reduce the risk of failure when trying to create a startup. The process described in this dissertation describes how to evaluate the key elements of an idea and conduct a series of interviews with potential customers to find evidence that supports pursing that idea further, challenge the startup team to change some aspect of the idea, or drop it altogether. Once the startup team has found a problem, as well as a solution to that problem, this dissertation describes an approach creating that solution. Then this dissertation describes an approach for critically evaluating the foundational elements of the problem and the solution. The goal for a critical evaluation is to identify additional foundational elements which relate to the product that may increase its value or decrease the risk of product failure.

Smart Specialisation Strategy (S3) has recently attracted the attention of many scholars, pundits and policy makers involved in regional studies, as a new industrial policy able to fill the gap between the weak capacity of Europe to innovate in comparison to its strong academic base and research institutions. S3 is described as a policy aimed to encourage structural changes, through the generation of new domains of opportunities, according to the strengths and potentialities of each region and therefore with a “place-based” outlook. Its primary element of novelty, in comparison to the previous policy approaches, is constituted by the Entrepreneurial Discovery Process (EDP), which represents the modality among institutions, firms, R&D centres, universities, through which the direction(s) of the structural change is organised. To study S3, this Ph.D. thesis focuses on two pillars considered central to understand its rationales: relatedness and entrepreneurship. On one hand, the idea of relatedness is useful to understand the economic structure of a territory and its evolution through its network of connections, outlining possible areas of future development. On the other hand, entrepreneurship, somehow a missing dimension of S3, can be considered as part of the process of opportunity scanning to “challenge” inefficiencies of the society through new models of production and consumption, proactiveness of institutions, business development strategies of firms or cultural mindset of people. The aim of the thesis is to explore this relatedness-entrepreneurship relationship within S3, using a multi-level framework of analysis able to integrate the different aspects of the two concepts, providing theoretical and empirical advancements. The thesis is structured as follows: a general introduction on S3, three papers, which analyse Italy, focusing on the case of Tuscany and some final conclusions that sum up the findings of the papers and provide some further policy insights. The content of the three papers is reported hereinafter. In the first paper the analysis is conducted in the Italian provinces defining entrepreneurship as the creation of a new business and relatedness as one of the principal mechanisms that could explain the origin of innovation in connection with a given territorial knowledge base. The distinctiveness of this first paper seeds in the study of this relationship across individual industries, computing separate measures of external and internal relatedness across 27 sectors (among manufacturing and KIBS). The results suggest a broader and positive impact of external relatedness on the concentration of new firms at the territorial level in comparison to the impact of internal relatedness. The implications suggest that Knowledge Spillover Theory of Entrepreneurship can be included in the cognitive framework of S3 (newborns as expression of knowledge exchanged at the local level) and that innovation policies aimed to promote path creation should consider existent strengths of the territories. The second paper studies the EDP, integrating the concept of relatedness, useful in the initial phases of design and scoping, with the one of institutional entrepreneurship as an expression of the impact of agency in the micro-dynamics that rule the final outcome of innovation policies. This framework is applied to the case of Tuscany, using a mixed methodology. As a first picture of proximity connections between sectors of Tuscany, an original computation of the “Industry Space” of Tuscany is realised (using the methodology of Hidalgo et al., 2007). Then the Technological Districts’ managers and/or coordinators are interviewed, as a sort of fact checking with the Industry Space results, to understand how they define their planning strategies and through which mechanisms they integrate knowledge and combine firms and R&D specialities. Results confirm the necessity to integrate the two concepts to obtain a more realistic “policy orientation map”, and the broader horizon released by relatedness if deeply analysed with case studies at a micro-level and if directly discussed with some central agents embedded in the regional network of proximities. The third paper studies the entrepreneurial styles (as real business men) and their ways of integrating and combining knowledge, adopting a micro interpretation on the concept of relatedness. The paper aims to identify what role can play these entrepreneurial figures as fundamental “micro pieces” in the scanning process of future opportunities of regional transformation promoted by S3. The methodology adopts a qualitative approach, using semi-structured interviews administered to a selected set of 24 entrepreneurs in Tuscany. The sample of the entrepreneurs, selected with a purposeful criterion, has been built thanks to the help of key informants. The gathered data are codified with the help of Gioia methodology, in order to derive some characteristics of the entrepreneur and the firms to describe some “emerging properties”. Then, a ladder of entrepreneurial typologies, able to group the specific characteristics derived from the interviews, is proposed. Results suggest a “distributed technology transfer model” as a complementary bottom up strategy to converge towards a new cyber-manufacturing regime of production.

Automatic Root Cause Analysis (RCA) systems aim to streamline the process of identifying the underlying cause of software failures in complex cloud-native environments. These systems employ graph-like structures to represent causal relationships between different components of a software application. These relationships are typically learned through performance and resource utilization metrics of the microservices in the system. To accomplish this objective, numerous RCA systems utilize statistical algorithms, specifically those falling under the category of causal discovery. These algorithms have demonstrated their utility not only in RCA systems but also in a wide range of other domains and applications. Nonetheless, there exists a research gap in the exploration of the feasibility and efficacy of multivariate time series causal discovery algorithms for deriving causal graphs within a microservice framework. By harnessing metric time series data from Prometheus and applying these algorithms, we aim to shed light on their performance in a cloudnative environment. Furthermore, we have introduced an adaptation in the form of an ensemble causal discovery algorithm. Our experimentation with this ensemble approach, conducted on datasets with known causal relationships, unequivocally demonstrates its potential in enhancing the precision of detected causal connections. Notably, our ultimate objective was to ascertain reliable causal relationships within Ericsson’s cloud-native system ’X,’ where the ground truth is unavailable. The ensemble causal discovery approach triumphs over the limitations of employing individual causal discovery algorithms, significantly augmenting confidence in the unveiled causal relationships. As a practical illustration of the utility of the ensemble causal discovery techniques, we have delved into the domain of anomaly detection. By leveraging causal graphs within our study, we have successfully applied this technique to anomaly detection within the Ericsson system. / System för automatisk rotorsaksanalys (RCA) syftar till att effektivisera process för att identifiera den underliggande orsaken till programvarufel i komplexa molnbaserade miljöer. Dessa system använder grafliknande strukturer att representera orsakssamband mellan olika komponenter i en mjukvaruapplikation. Dessa relationer lär man sig vanligtvis genom prestanda och resursutnyttjande mätvärden för mikrotjänsterna i systemet. För att uppnå detta mål använder många RCAsystem statistiska algoritmer, särskilt de som faller under kategorin orsaksupptäckt. Dessa algoritmer har visat att de inte är användbara endast i RCA-system men även inom en lång rad andra domäner och applikationer. Icke desto mindre finns det en forskningslucka i utforskningen av genomförbarhet och effektivitet av orsaksupptäckt av multivariat tidsserie algoritmer för att härleda kausala grafer inom ett mikrotjänstramverk. Genom att utnyttja metriska tidsseriedata från Prometheus och tillämpa Dessa algoritmer strävar vi efter att belysa deras prestanda i ett moln- inhemsk miljö. Dessutom har vi infört en anpassning i formen av en ensemble kausal upptäcktsalgoritm. Vårt experiment med denna ensemblemetod, utförd på datauppsättningar med kända orsakssamband relationer, visar otvetydigt sin potential för att förbättra precisionen hos upptäckta orsakssamband. Särskilt vår ultimata Målet var att fastställa tillförlitliga orsakssamband inom Ericssons molnbaserade systemet ’X’, där grundsanningen inte är tillgänglig. De ensemble kausal discovery approach segrar över begränsningarna av att använda individuella kausala upptäcktsalgoritmer, avsevärt öka förtroendet för de avslöjade orsakssambanden. Som en praktisk illustration av nyttan av ensemblens kausal upptäcktstekniker har vi fördjupat oss i anomalidomänen upptäckt. Genom att utnyttja kausala grafer inom vår studie har vi framgångsrikt tillämpat denna teknik för att detektera anomali inom Ericsson system