| 1. |
- Babu, Md Abu Ahammed, 1994, et al.
(författare)
-
Impact of Image Data Splitting on the Performance of Automotive Perception Systems
- 2024
-
Ingår i: Lecture Notes in Business Information Processing. - 1865-1356 .- 1865-1348. ; 505 LNBIP, s. 91-111
-
Konferensbidrag (refereegranskat)abstract
- Context: Training image recognition systems is one of the crucial elements of the AI Engineering process in general and for automotive systems in particular. The quality of data and the training process can have a profound impact on the quality, performance, and safety of automotive software. Objective: Splitting data between train and test sets is one of the crucial elements in this process as it can determine both how well the system learns and generalizes to new data. Typical data splits take into consideration either randomness or timeliness of data points. However, in image recognition systems, the similarity of images is of equal importance. Methods: In this computational experiment, we study the impact of six data-splitting techniques. We use an industrial dataset with high-definition color images of driving sequences to train a YOLOv7 network. Results: The mean average precision (mAP) was 0.943 and 0.841 when the similarity-based and the frame-based splitting techniques were applied, respectively. However, the object-based splitting technique produces the worst mAP score (0.118). Conclusion: There are significant differences in the performance of object detection methods when applying different data-splitting techniques. The most positive results are the random selections, whereas the most objective ones are splits based on sequences that represent different geographical locations.
|
|
| 2. |
- Havers, Bastian, 1991, et al.
(författare)
-
DRIVEN: a Framework for Efficient Data Retrieval and Clustering in Vehicular Networks
- 2019
-
Ingår i: Proceedings - International Conference on Data Engineering. - 1084-4627. ; 2019-April, s. 1850-1861
-
Konferensbidrag (refereegranskat)abstract
- Applications for adaptive (sometimes also called smart) Cyber-Physical Systems are blossoming thanks to the large volumes of data, sensed in a continuous fashion, in large distributed systems. The benefits of these applications come nonetheless with a price: the need for jointly addressing challenges in efficient data communication and analysis (among others). The goal of the DRIVEN framework, presented here, is to address these challenges for a data gathering and distance-based clustering tool in the context of vehicular networks. Because of the limited communication bandwidth (compared to the volume of sensed data) of vehicular networks and the monetary costs of data transmission, the intuition behind DRIVEN is to avoid gathering the data to be clustered in a raw format from each vehicle, but rather to allow for a streaming-based error-bounded approximation, through Piecewise Linear Approximation, to compress the volumes of data to be gathered. At the same time, rather than relying on a batch-based clustering algorithm that requires all the data to be first gathered (and then clustered), DRIVEN relies on and extends a streaming-based clustering algorithm that leverages the inherent ordering of the spatial and temporal data being collected, to perform the clustering in an online fashion, while data is being retrieved. As we show, based on our prototype implementation using Apache Flink and our evaluation with real-world data such as GPS and LiDAR, the accuracy loss for the clustering performed on the reconstructed data can be small, even when the raw data is compressed to 10- 35% of its original size, and the transferring of data itself can be completed in up to one-tenth of the duration observed when gathering raw data.
|
|
| 3. |
- Havers, Bastian, 1991, et al.
(författare)
-
DRIVEN: A framework for efficient Data Retrieval and clustering in Vehicular Networks
- 2020
-
Ingår i: Future Generation Computer Systems. - : Elsevier BV. - 0167-739X. ; 107, s. 1-17
-
Tidskriftsartikel (refereegranskat)abstract
- The growing interest in data analysis applications for Cyber–Physical Systems stems from the large amounts of data such large distributed systems sense in a continuous fashion. A key research question in this context is how to jointly address the efficiency and effectiveness challenges of such data analysis applications. DRIVEN proposes a way to jointly address these challenges for a data gathering and distance-based clustering tool in the context of vehicular networks. To cope with the limited communication bandwidth (compared to the sensed data volume) of vehicular networks and data transmission's monetary costs, DRIVEN avoids gathering raw data from vehicles, but rather relies on a streaming-based and error-bounded approximation, through Piecewise Linear Approximation (PLA), to compress the volumes of gathered data. Moreover, a streaming-based approach is also used to cluster the collected data (once the latter is reconstructed from its PLA-approximated form). DRIVEN's clustering algorithm leverages the inherent ordering of the spatial and temporal data being collected to perform clustering in an online fashion, while data is being retrieved. As we show, based on our prototype implementation using Apache Flink and thorough evaluation with real-world data such as GPS, LiDAR and other vehicular signals, the accuracy loss for the clustering performed on the gathered approximated data can be small (below 10%), even when the raw data is compressed to 5-35% of its original size, and the transferring of historical data itself can be completed in up to one-tenth of the duration observed when gathering raw data.
|
|
| 4. |
- Havers, Bastian, 1991, et al.
(författare)
-
Proposing a framework for evaluating learning strategies in vehicular CPSs
- 2022
-
Ingår i: Middleware 2022 Industrial Track - Proceedings of the 23rd International Middleware Conference Industrial Track, Part of Middleware 2022. - New York, NY, USA : ACM. - 9781450399173 ; , s. 22-28
-
Konferensbidrag (refereegranskat)abstract
- Highly-connected Vehicular Cyber-Physical Systems (VCPSs) offer manifold opportunities for distributing learning across the contained vehicles, road-side units and servers. However, simulating and evaluating particular distributed learning schemes poses a difficult problem in requiring realistic modeling of the vehicular fleet, communication, and the learning itself. In this work, we postulate a set of requirements for a framework simulating a complete learning workflow in a VCPS, and propose a modular architecture for it. Using a prototype implementation, we show with an example experiment the capabilities the proposed framework delivers for evaluating novel learning schemes in custom scenarios.
|
|
| 5. |
- Koppisetty, Ashok Krishna Chaitanya, 1982
(författare)
-
Application and Development of Computational Methods for Structure-based Drug Discovery
- 2014
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Researchers involved in drug discovery need tools to support their decisions regarding which of many available chemical entities could be developed into potent drug candidates. The qualitative and quantitative results from computational docking methods constitute the basis for structure-based rational design of inhibitors in the early stages of drug discovery. Research contributions related to application and development of such computational methods are presented here.Initially, existing computational methods were used to obtain insights into the functionality of human enzyme Arylsulfatase A and the adhesion of Norovirus VA387 to human ABO blood group saccharides. The computational results were validated either with existing experimental data or with data from new experiments, and have significant importance for rational design of inhibitors for Arylsulfatase A and Norovirus VA387. These studies motivated the development and evaluation of new computational methods for estimating binding energies, and their thermodynamic components, of protein-ligand complexes. Support vector machine based scoring functions were developed to estimate the binding energies of the protein-ligand interactions including their enthalpy and entropy components with significant accuracy. The estimations by the reported scoring functions are seen to outperform the existing scoring functions in benchmarks with protein-ligand complexes from the PDBbind database. Methods based on expanded ensemble molecular dynamics simulations were explored for the first time for estimating the binding energies and their thermodynamic components. Binding energies were estimated for interactions of plant lectin hevein with its carbohydrate ligands, giving results that are in good agreement with the existing experimental binding data. These methodological developments related to estimating binding energies and their thermodynamic components should be valuable for profiling the binding characteristics of lead compounds that could be developed to potential drug candidates.
|
|
| 6. |
- Koppisetty, Ashok Krishna Chaitanya, 1982, et al.
(författare)
-
Binding energy calculations for hevein-carbohydrate interactions using expanded ensemble molecular dynamics simulations
- 2015
-
Ingår i: Journal of Computer-Aided Molecular Design. - : Springer Science and Business Media LLC. - 0920-654X .- 1573-4951. ; 29:1, s. 13-21
-
Tidskriftsartikel (refereegranskat)abstract
- Accurate estimation of protein-carbohydrate binding energies using computational methods is a challenging task. Here we report the use of expanded ensemble molecular dynamics (EEMD) simulation with double decoupling for estimation of binding energies of hevein, a plant lectin with its monosaccharide and disaccharide ligands GlcNAc and (GlcNAc)(2), respectively. In addition to the binding energies, enthalpy and entropy components of the binding energy are also calculated. The estimated binding energies for the hevein-carbohydrate interactions are within the range of +/- 0.5 kcal of the previously reported experimental binding data. For comparison, binding energies were also estimated using thermodynamic integration, molecular dynamics end point calculations (MM/GBSA) and the expanded ensemble methodology is seen to be more accurate. To our knowledge, the method of EEMD simulations has not been previously reported for estimating biomolecular binding energies.
|
|
| 7. |
- Koppisetty, Ashok Krishna Chaitanya, 1982, et al.
(författare)
-
Computation of Binding Energies Including Their Enthalpy and Entropy Components for Protein-Ligand Complexes Using Support Vector Machines
- 2013
-
Ingår i: Journal of Chemical Information and Modeling. - : American Chemical Society (ACS). - 1549-9596 .- 1549-960X .- 0095-2338 .- 1520-5142. ; 53:10, s. 2559-2570
-
Tidskriftsartikel (refereegranskat)abstract
- Computing binding energies of protein-ligand complexes including their enthalpy and entropy terms by means of computational methods is an appealing approach for selecting initial hits and for further optimization in early stages of drug discovery. Despite the importance, computational predictions of thermodynamic components have evaded attention and reasonable solutions. In this study, support vector machines are used for developing scoring functions to compute binding energies and their enthalpy and entropy components of protein-ligand complexes. The binding energies computed from our newly derived scoring functions have better Pearson's correlation coefficients with experimental data than previously reported scoring functions in benchmarks for protein-ligand complexes from the PDBBind database. The protein-ligand complexes with binding energies dominated by enthalpy or entropy term could be qualitatively classified by the newly derived scoring functions with high accuracy. Furthermore, it is found that the inclusion of comprehensive descriptors based on ligand properties in the scoring functions improved the accuracy of classification as well as the prediction of binding energies including their thermodynamic components. The prediction of binding energies including the enthalpy and entropy components using the support vector machine based scoring functions should be of value in the drug discovery process.
|
|
| 8. |
- Koppisetty, Ashok Krishna Chaitanya, 1982
(författare)
-
Computational Prediction and Analysis of Protein-Carbohydrate and Protein-Protein Interactions
- 2010
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The structure and energetics of protein-carbohydrate and protein-protein interactions are of great interest due to their importance in many biological phenomena. Two cases of protein-carbohydrate interactions and one case of protein-protein interactions were predicted and analyzed using existing computational methods. Computational methods are shown to provide valuable insights for problems which are difficult to approach with experimental methods. Addressing the absence of crystallographic data, molecular docking methods are used to study the interactions of the enzyme Arylsulfatase A (ASA) with its natural sulfoglycolipid substrates and the activator protein saposin B (Sap B). Some of the predicted interactions of ASA with its natural sulfoglycolipid substrates were verified through mutational studies. The preliminary results of ASA interaction with Sap B were analyzed with existing clinical data on mutations of ASA. These results have implications in designing experiments involving ASA, Sap B and its substrates. The second case concerns binding of norovirus VA387 with a variety of histo-blood group ABO active carbohydrate structures. Using an available crystal structure of the VA387 capsid protein with B-trisaccharide, detailed interactions for 11 carbohydrate structures were predicted using molecular dynamics and binding affinity scoring. The predicted interactions stand in good agreement with the existing mutational data on the binding of VA387 to carbohydrates. The results could be useful in structure-based drug design of adhesion inhibitors for noroviruses. In the scoring functions of current molecular docking methods, it is seen that entropic contributions are neglected. Methodological development for predicting entropic contributions in binding are planned as future studies.
|
|
| 9. |
- Koppisetty, Ashok Krishna Chaitanya, 1982, et al.
(författare)
-
Computational studies on the interaction of ABO-active saccharides with the norovirus VA387 capsid protein can explain experimental binding data
- 2010
-
Ingår i: Journal of Computer-Aided Molecular Design. - : Springer Science and Business Media LLC. - 0920-654X .- 1573-4951. ; 24:5, s. 423-431
-
Tidskriftsartikel (refereegranskat)abstract
- Norovirus strains are known to cause recurring epidemics of winter vomiting disease. The crystal structure of the capsid protein of VA387, a representative of the clinically important GII.4 genocluster, was recently solved in complex with histo-blood group A- and B-trisaccharides. However, the VA387 strain is known to bind also to other natural carbohydrates for which detailed structural information of the complexes is not available. In this study we have computationally explored the fit of the VA387 with a set of naturally occurring carbohydrate ligands containing a terminal alpha 1,2-linked fucose. MD simulations both with explicit and implicit solvent models indicate that type 1 and 3 extensions of the ABO-determinant including ALe(b) and BLe(b) pentasaccharides can be well accommodated in the site. Scoring with Glide XP indicates that the downstream extensions of the ABO-determinants give an increase in binding strength, although the alpha 1,2-linked fucose is the single strongest interacting residue. An error was discovered in the geometry of the GalNAc-Gal moiety of the published crystal structure of the A-trisaccharide/VA387 complex. The present modeling of the complexes with histo-blood group A-active structures shows some contacts which provide insight into mutational data, explaining the involvement of I389 and Q331. Our results can be applicable in structure-based design of adhesion inhibitors of noroviruses.
|
|
| 10. |
- Nasir, Waqas, et al.
(författare)
-
Lewis histo-blood group alpha1,3/alpha1,4 fucose residues may both mediate binding to GII.4 noroviruses
- 2012
-
Ingår i: Glycobiology. - : Oxford University Press (OUP). - 0959-6658 .- 1460-2423. ; 22:9, s. 1163-72
-
Tidskriftsartikel (refereegranskat)abstract
- Human noroviruses cause recurrent epidemics of gastroenteritis known to be dominated by the clinically important GII.4 genotype which recognizes human Secretor gene-dependent ABH histo-blood group antigens (HBGAs) as attachment factors. There is increasing evidence that GII.4 noroviruses have undergone evolutionary changes to recognize Lewis antigens and non-Secretor saliva. In this study, we have investigated the possibilities of the Lewis alpha1,3/alpha1,4 fucoses as mediators of binding of GII.4 noroviruses to Lewis antigens. The study was carried out using molecular dynamics simulations of Lewis type-1 and type-2 chain HBGAs in complex with VA387 P domain dimers in explicit water. Based on the computer simulations, we suggest the possibility of two receptor binding modes for Lewis HBGAs: the "Secretor pose" with the Secretor Fucalpha1,2 in the binding site and the "Lewis pose" with the Lewis Fucalpha1,3/alpha1,4 residues in the binding site. This was further supported by an extensive GlyVicinity analysis of the Protein Data Bank with respect to the occurrence of the Lewis and Secretor poses in complexes of Lewis antigens with lectins and antibodies as well as GII norovirus strains. The Lewis pose can also explain the interactions of GII.4 norovirus strains with Le(x) and SLe(x) structures. Moreover, the present model suggests binding of complex branched polysaccharides, with the Lewis antigens at the nonreducing end, to P domain dimers of GII.4 strains. Our results are relevant for understanding the evolution of norovirus binding specificities and for in silico design of future antiviral therapeutics.
|
|
