Indian Institute of Science Bangalore
etd@IISc Electronic Theses and Dissertations at Indian Institute of ScienceNot a member yet
6204 research outputs found
Sort by
Identifying Bacterial Species, Detecting Antibiotic Resistance and Susceptibilities in Clinical Samples Using Raman Spectroscopy
No full textOver the past few decades, Raman spectroscopy has found enormous applications in biology and medicine. Raman micro-spectroscopy is highly sensitive to the structure and composition of molecules and their surrounding moieties present in a sample. This technique is label-free and non-invasive and offers a significant advantage over other methods involving minimal sample preparation and water interference. Raman spectroscopy combined with various chemometrics methods can provide molecular insights into several biochemical processes. This thesis discusses the application of Raman spectroscopy in identifying bacterial species, determining antibiotic susceptibility in clinical samples, and detecting phenotypic antibiotic resistance.
The thesis begins with a brief introduction to the principle and instrumentation of Raman spectroscopy (Chapter 1). It provides a brief literature review of the applications of Raman spectroscopy in studying biological molecules. Chapter 2 of the thesis discusses the various Raman spectral data analysis approaches.
Herbicides are ubiquitous in modern society and often co-occur with antibiotics in runways, sewage, and water bodies (Chapter 3). The herbicides used in this study are 2,4 D and Glyphosate, as these are the most widely used herbicides in agriculture worldwide. The study uses a group of isogenic mutants of E. coli, ∆lon, and ∆acrB, displaying different antibiotic susceptibilities. Combined with supervised machine learning methods, Raman spectroscopy could efficiently track the emergence of antibiotic resistance and confirm that the induction of antibiotic resistance is AcrB-dependent.
Identifying bacterial species from clinical samples is often challenging. The current gold standard for identifying bacteria is time-intensive, laborious, and may provide incorrect results if the sample matrix is not entirely removed (Chapter 4). This part of the thesis demonstrates Raman spectroscopy-based identification of pathogens from clinical samples using deep learning at the single-cell level, especially for the ESKAPE class of pathogens.
There has been an exponential rise in the emergence of antibiotic-resistant bacteria in the past few decades. Conventional methods for determining antibiotic susceptibilities do not provide information about antibiotic susceptibilities and fail to detect unknown resistant genes (Chapter 5). This part of the thesis shows the potential of Raman spectroscopy in differentiating various antibiotic-treated and untreated clinical isolates of Escherichia coli, Acinetobacter baumannii, and Enterobacter species into resistant and sensitive strains. Finally, the work done in the thesis and the future scope of the work has been briefed (Chapter 6). Overall, the thesis describes the applicability of Raman spectroscopy ranging from basic research to clinics
Resolvin D1-loaded Liposomal Nanocarriers for the Treatment of Osteoarthritis
No full textOsteoarthritis (OA) is the most common disorder of the joints and currently affects >300 million patients worldwide. In this disease, the articulating cartilage undergoes progressive damage which leads to severe pain and loss of function. Current treatment for OA does not target the etiology of the disease but instead focuses only on alleviating pain by administration of steroidal analgesics. Steroidal analgesics are not ideal for the long-term treatment of the disease because they generate chronic chondrotoxicity and cartilage thinning. Eventually, as the disease progresses towards severe stages, the patients have to undergo a highly invasive and costly joint replacement surgery. Currently, there is no disease-modifying OA drug (DMOAD) approved for human use.
OA-associated damage is driven by chronic low-grade inflammation, and several cells are involved in the cross-talk that contributes to the inflammation. Targeting the OA-related chronic inflammation using antagonists and antibodies against inflammatory cytokines is a viable strategy to treat OA and has shown promise in pre-clinical trials. However, previous attempts to deliver anti-inflammatory molecules to the site of damage did not translate into therapy for humans. The failure of these trials can be attributed to the fast clearance of the small molecules by synovial vasculature and of the macromolecules by the lymphatics, which results in short half-lives in the joint (1- 4h) for commonly used steroids. The limiting in vivo half-lives of small-molecule drugs can be extended by encapsulating the molecules in particle-based systems. Liposomes are ideal for efficient delivery because of their biodegradability, low toxicity, stability, flexible synthesis methods, and ability to incorporate versatile cargo.
Specialized proresolution mediators (SPMs) are potent molecules that actively reduce inflammatory factors from the site of damage in a process called resolution of inflammation. Chronic inflammatory diseases, including OA, are known to have impaired inflammation resolution pathways. Exogenously supplied SPMs like Resolvin D1 (RvD1) reduce the severity of OA, but strategies involving the direct IA injection of such molecules require administration of doses before OA-causing injury has taken place. In this study, we generate a liposomal-RvD1(lipo-RvD1) formulation to increase the bioavailability of the molecule in the pathology-affected joint. Liposomes were actively loaded with RvD1 using a pH gradient strategy. This formulation was stable and acted as a source of RvD1 for ~11 days in vitro in near-physiological conditions.
We evaluated the prophylactic efficacy of our formulation in the widely used, Destabilization of Medial Meniscus (DMM) mouse model of OA. Our histology results show that only lipo-RvD1 treated joints reduced the severity of OA, whereas free RvD1 did not have any protective effect on the joint. We conducted the same experiment in an obesity-related OA model in mice to reproduce the contribution of obesity to the pathology of the disorder. The pathology of obesity-related OA is much more severe due to the contribution of adipose tissue to OA-associated inflammation. In this model also, only the lipo-RvD1 treated joints had shown efficacy whereas free-RvD1 had no protective effect on the joint. The results from the latter model established the ability of lipo-RvD1 to treat diseases in diverse stages of progression.
Most OA patients report to the clinics in the later stages of the disease. Thus, there is a requirement for a good therapeutic agent for the treatment of OA. In the therapeutic study, we administered the formulation to the animals after significant damage is known to have ensued. Lipo-RvD1 treated joints showed much healthier articular cartilage compared to free-RvD1 and surgery-only joints in both mouse models.
We observed that administration of lipo-RvD1 was more effective in alleviating the allodynia than DMM mice and free RvD1 injected mice in the therapeutic regimen of administration. Our results show that the sustained presence of RvD1 in the affected knee joint helps alleviate OA-associated pain
Graph Neural Networks with Parallel Local Neighborhood Aggregations
No full textGraph neural networks (GNNs) have become very popular for processing and analyzing graph-structured data in the last few years. Using message passing as their basic building blocks that aggregate information from neighborhoods, GNN architectures learn low-dimensional graph-level or node-level embeddings useful for several downstream machine learning tasks. In this thesis, we focus on GNN architectures that perform parallel neighborhood aggregations (in short, referred to as PA-GNNs) for two tasks, namely, graph classification and link prediction. Such architectures have a natural advantage of reduced training and inference time as neighborhood aggregation is done before training, unlike GNNs that perform the neighborhood aggregation sequentially (referred to as SA-GNNs) during training. Thus, the runtime of SA-GNNs depends on the number of edges in the graph.
In the first part of the thesis, we propose a generic model for GNNs with parallel neighborhood aggregation and theoretically characterize the discriminative power of PA-GNN models to discriminate between two non-isomorphic graphs. We provide conditions under which they are provably powerful as the well-known Weisfeiler-Lehman graph isomorphism test. We then specialize the generic PA-GNN model and propose a GNN architecture, which we call SPIN: simple and parallel graph isomorphism network. Next to the theoretical characterization of the developed PA-GNN model, we also present numerical experiments on a diverse variety of real-world benchmark graph classification datasets related to social networks, chemical molecular data, and brain networks. The proposed model achieves state-of-the-art performance with reduced training and inference time.
In the second part of the thesis, we propose a computational method for drug repurposing by presenting a deep learning model that captures the complex interactions between the drugs, diseases, genes, and anatomies in a large-scale interactome with over 1.4 million connections. Specifically, we propose a PA-GNN based drug repurposing architecture, which we call GDRnet, to screen a large drug database of clinically approved drugs and predict potential drugs that can be repurposed for novel diseases. GNN-based machine learning models are a natural choice for computational drug repurposing because of their ability to capture the underlying structural information in such complex biological networks. While the proposed PA-GNN architecture is computationally attractive, we present results from numerical experiments to show the efficacy of GNNs for drug repurposing. We also provide numerical experimental results on drug repurposing for coronavirus diseases (including COVID-19), where many of the drugs predicted by the proposed model are considered as the mainstay treatment
Fast Algorithms for Max Cut on Geometric Intersection Graphs
No full textIn this work, we design fast algorithms for max cut on geometric intersection graphs. In the maximum cut (a.k.a., max cut) problem, the input is an undirected graph, and the goal is to partition the vertex set into two disjoint sets such that the number of edges having their endpoints in different sets is maximized. In a geometric intersection graph, given a collection of n geometric objects as input (say unit intervals or unit squares), each object corresponds to a vertex and there is an edge between two vertices if and only if the corresponding objects intersect. Note that the edge-set of the graph is not explicitly given as input.
A simple O(n) time algorithm places each object randomly into one of the two sets to achieve a 0.5 approximation. Our goal is to design near-linear time algorithms (in terms of n) and beat the 0.5 approximation barrier. We obtain the following results:
• An O(nlogn) time algorithm with an approximation factor of 2/3 for unit interval in- tersection graphs. We decompose the unit intervals into several cliques and based on the number of edges between “adjacent” cliques, we choose an appropriate partitioning strategy.
• An O(nlogn) time algorithm with an approximation factor of 7/13 for unit square in- tersection graphs. We iteratively find the deepest point among the unit squares and then remove the squares containing the deepest point. To obtain a fast algorithm, we design a data structure which can efficiently answer deepest point queries and also, efficiently handle deletion of unit squares.
• An exact and fast algorithm for laminar geometric intersection graphs. Our algorithm uses a simple greedy strategy; however, proving its correctness requires non-trivial ideas. A fast algorithm is obtained by combining the properties of laminar objects with range searching data structures
Solution structures and interaction studies of mycobacterial MazEF6 toxin-antitoxin system
No full textToxin-Antitoxin (TA) systems are stress response systems found on plasmids and chromosomes
of bacteria and archaea. They are involved in important functions such as
plasmid maintenance, programmed cell death, persister cell formation, phage abortive
infection, bacteriostasis and formation of antibiotic -resistant population. The H37Rv
strain of Mtb encodes an exceptionally high number ( 90) of TA systems, while the
non-pathogenic mycobacterial species, M. smegmatis encodes only 5 TA systems which
is relatively fast growing. Majority of the TA systems encoded in the Mtb H37Rv genome
are of Type II, MazEF system being a prominent member. The studies have shown that MazE6 and nMazE6 bind to the cognate operator DNA harboring
the -10 promoter region (TATACT) with high affinity. The high resolution NMR
studies have identified the specific residues neccessary for this binding. Furthermore, it
has been shown that the MazE6 does not bind to the non-cognate mazEF9 operator
DNA. It has been shown that specificty of interaction lies in the operator DNA sequence
as well. Mutation of the operator DNA sequence at the NNN position in TANNNT
(TATACT) sequence resulted in weakening of the interaction by 100 fold. Thus, the
structures of mycobacterial protein MazE6 and MazF6 from MazEF6 TA complex and
function of MazE6 as an antitoxin and as a transcription regulator has been established
through this projec
Development of Microneedles for Transdermal and Intradermal Drug Delivery
No full textTransdermal drug delivery has an enormous potential for painless therapeutics. Many drugs designed for systemic action are more efficient when delivered through the skin compared to oral administration as their degradation in the digestive tract is prevented. Vaccines delivered in the skin(Intradermal) also require much less dose volume compared to conventional intramuscular route for the same efficacy. Since transdermal delivery is painless, it can present a great advantage to people dependent on daily injection therapies like insulin. However, molecular weight of topically applied drugs that can penetrate through the skin is limited to 400 Da. The main challenge in achieving a successful drug delivery through the skin involves designing strategies to overcome permeation resistance offered by the top skin layer called stratum corneum (SC). Tiny, sharp projections designed to penetrate SC, called ”Microneedles” are very effective in enhancing this molecular weight limit.
Any new product development cycle involves a sequential process of design, identification of feasible manufacturing processes, development of a first successful prototype, pilot batch fabrication, testing and mass manufacturing after successful testing of the pilot batch. But for a specialized segment like microneedle manufacturing, several iterations may be required between design and testing phase. For the microneedle fabrication techniques reported in the literature, involving use of sophisticated technologies and equipment, the costs involved in iterating designs are extremely high. Major challenge in enhancing the utility of microneedles for widespread applications lies in cutting down the cost of fabrication methods and increasing the flexibility of tweaking geometric parameters.
The present thesis discusses methods developed for fabrication of microneedles keeping design simplicity, flexibility and multiplicity in perspective. In first part of the thesis lithography methods were explored for the fabrication of solid SU8 microneedles and their limitations are discussed. In later parts, a syringe retrofit hollow microneedle device for generic drug delivery, made using stainless steel needles is discussed. The device has been extensively tested on animals and its utility has been verified. In the final part, a variable aspect ratio molding platform for fabrication of solid and dissolvable microneedles is discussed. This unique platform offers the capability of fabricating series of needle patches with heights ranging from as low as 200-300 μm to as high as1800-2000 μm with resolution of 50 μm. The platform’s capability of fabricating hollow microneedle patches using drop casting method and a metered dosing package for them is also discussed. Overall, the work presented in this thesis takes a step towards development of a low-cost product with a potential of productionization
Gaze Estimation in the Wild - Models, Datasets and Usability
No full textHuman eye gaze estimation research has numerous applications in diverse fields from Human Computer Interaction (HCI) to aviation. Non-intrusive video oculography based methods are classified into two categories based on their input image type. Infra-red illuminator-based approaches report higher accuracy and are available commercially. Appearance-based gaze estimation systems, on the other hand, do not require any dedicated hardware and relies on RGB camera image, hence can be deployed at a larger scale. Existing appearance-based gaze estimation approaches are either less accurate or too bulky to be deployed for interaction purposes. I proposed two gaze estimation models based on attention and difference mechanisms. The proposed AGE-Net model achieved state of the art accuracy on all publicly available datasets. Further, the proposed models contain less memory footprint and can be deployed for real-time interaction. Next, analysis of existing appearance-based gaze estimation datasets revealed that their head pose angles and illumination variations are limited. Further, no dataset or evaluation talks about the precision of gaze estimates from the models proposed so far. In this direction, I proposed PARKS-Gaze, a precision-focused gaze estimation dataset which was recorded in daily life conditions and captured wider head pose and illumination levels than the existing datasets like MPIIGaze and GazeCapture. Even though IR-based systems are commercially available, their performance limitations were not well studied. This hinders their deployment in wide range of scenarios. To this end, a feasibility assessment of a commercially available IR-based gaze estimation wearable eye tracker in actual flying conditions was conducted. Experiments carried out in two fighter aircrafts revealed that the eye tracker failed to provide consistent gaze estimates. Similar findings were observed in terms of detection rate in the studies conducted in laboratory conditions as well. I proposed two hypotheses about the failure modes of commercial eye gaze tracker in flying environment and corroborated them with in-flight eye tracking data. Finally, an indigenous helmet-mounted eye tracker was designed to further extend the gaze estimation capabilities. The proposed helmet-mounted eye tracker along with a deep-learning based gaze estimation method achieved on-par accuracy, superior precision error and consistent gaze estimates compared to a commercial eye tracker. This thesis concludes by presenting the applications developed using the proposed systems and their usability in automotive, aviation and assistive technology domains
Towards Design, Fabrication, Packaging, Integration, and Characterization of High-Performance MEMS Gyroscopes
No full textMEMS vibratory gyroscopes are used to measure the angular rate of a body by
sensing the Coriolis force induced motion of the sensing element vibrating in a rotating
frame of reference. MEMS gyroscopes find use in a wide variety of applications such as
smartphones, automobiles, navigation, bio medical instruments, industrial systems, etc.
A considerable amount of research has been done over the past two decades towards the
improvement of MEMS gyroscope performance in terms of its scale factor, resolution,
dynamic range, temperature sensitivity and bandwidth, with a goal of using them in
high performance applications. Industrial report published about a year ago predicted
the advent of inertial grade MEMS gyroscopes into the market by 2030. Towards this
goal, active research is being pursued across the world by solving various issues towards
the improvement of signal to noise ratio in MEMS gyroscopes stretching them into the
ultimate performance regime at par with their optical counterparts.
This work focuses on the complete development of a MEMS gyroscope suitable for
high-performance applications. This involves the thorough study of specifications, meticulous
design of the sensing structure to extract the required performance parameters,
incorporation of fabrication related non-idealities into the design, precise fabrication
of the sensor structure with tight tolerance, complete die-level characterization of the
sensor, vacuum sealed packaging and integration with electronics to extract the sensor
response to applied angular rates. The technique of sensitivity analysis is used to
optimize the design and the same is demonstrated on three different designs of MEMS
gyroscopes. DRIE process related non-idealities like the slanting and scalloping profiles
are incorporated into the modelling of gyroscope structures. Extensive simulations
are carried out to study the dependence of the etch profiles on different performance
criteria of the MEMS gyroscope. Two designs are finalized and carried forward for
fabrication. Combined wet and dry bulk micromachining technique is used to fabricate
the sensor chips. Two different fabrication process flows are optimized to realize the two different designs. Novel findings in the area of TMAH based wet bulk micromachining
for long duration etching and NH2OH added KOH based wet bulk micromachining at
low temperatures for faster etch rates are reported. DRIE is used for precise controlled
etching of the 100 μm device layer to realize the gyroscope structure. The fabricated
sensor chips are subsequently characterized mechanically to extract the drive and sense
resonant frequencies. A novel mechanical characterization set up for multiple sensor
chips is discussed. The experimental observations are compared with FEM simulation
values. Electrical characterization of the sensor chip involving C-V measurements and
drive voltage optimization is also reported. The packaging of the sensor chip is carried
out using an 84 pin quad flat leadless ceramic package. This package is sealed in vacuum
of a few mTorr. The sealed sensor package is then used for further electronic integration.
The electronic integration is carried out in two different ways. The sensor is
directly integrated with AD7746 24-bit capacitance-to-digital converter for extracting the
sense capacitance variation with angular rate. In the second method, a Trans-Impedance
Amplifier circuit and a Lock-in Amplifier are used for real time measurements of sensor
response to angular rate variation. A sensitivity of about 9.27 mV/(deg/s) is obtained
with a non-linearity of about 0.5% over the entire range of ±440 deg/s. Further, raw
data is continuously recorded for about 7 hrs with gyroscope in drive excitation mode
and without application of angular rate. This data is used to plot Allan deviation curve
which shows a bias instability of <5 deg/hr. The values of the specifications achieved,
point towards a high- performance gyroscope even with the sensor packaged separately.
The integration of the sensor chip with a C-V conversion ASIC is also briefly discussed
for hybrid packaging to achieve even better performance.
This thesis therefore presents some novel findings in the area of design, fabrication
and characterization of high-performance MEMS gyroscope
Construction of C-C Bonds by Photocatalysis via Radical Addition Cascade Cyclization (RACC): Synthesis of Heteroatom-Containing Small Molecules
No full textThe thesis presents photocatalytic methods for synthesizing small organic molecules at room temperature. The thesis is divided into four sections. Section A presents the general introduction to the thesis. Section B is presented as a single chapter, which describes an intramolecular ipso-arylative cyclization of aryl alkynoate with tricarbonyl compounds under photocatalytic conditions via radical addition cascade cyclization (RACC). The methodology could be extended to the arylpropiolamides and arylthioates as well. Section C is presented in two Chapters, Chapters 3 and 4. Chapter 3 describes a visible light-mediated difunctionalization of activated alkynes to form coumarin and spirolactones. The coumarins and spirolactones could be formed selectively by tuning the substituent present in the aryl alkynoate. The spirolactones synthesized in this method can be transformed into complex spirocyclic structures under a separate visible light-mediated reaction. Chapter 4 describes a visible light-mediated radical addition cascade cyclization (RACC) between maleimides and phenethylboronic acids to access fused tetrahydro-1H-benzo[e]isoindole-1,3(2H)-diones. The last section is presented as Chapter 5, which deals with a photocatalytic approach to the direct acylation of electron-deficient heteroarenes. The methodology could be used to synthesize biologically active small molecules
Understanding Nonhomologous DNA End Joining: Novel Insights Into Mechanism and Regulation
No full textThis study has provided new insights into the mechanistic details of NHEJ
and showed that DNA sequence around the break site can be a factor in determining which
proteins involved in NHEJ might be recruited with a higher efficiency. We have also generated
a KU70-mutant Nalm6 cell line (N6-Ku70-2) using CRISPR-Cas9 genome editing technology
and explored the differences between these cells and WT Nalm6 cells in terms of cell
proliferation and DNA repair potential. Finally, we have investigated the importance of
RNASEH2A/RNASEH2B expression in the maintenance of proper DSB repair and cellular
viability in T-ALL cells