| 1. |
- Ahlford, Annika, 1980-
(författare)
-
Applications of Four-Colour Fluorescent Primer Extension Technology for SNP Analysis and Discovery
- 2010
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Studies on genetic variation can reveal effects on traits and disease, both in humans and in model organisms. Good technology for the analysis of DNA sequence variations is critical. Currently the development towards assays for large-scale and parallel DNA sequencing and genotyping is progressing rapidly. Single base primer extension (SBE) is a robust reaction principle based on four-colour fluorescent terminating nucleotides to interrogate all four DNA nucleotides in a single reaction. In this thesis, SBE methods were applied to the analysis and discovery of single nucleotide polymorphism (SNP) in the model organism Drosophila melanogaster and in humans. The tag-array minisequencing system in a microarray format is convenient for intermediate sized genotyping projects. The system is scalable and flexible to adapt to specialized and novel applications. In Study I of the thesis a tool was established to automate quality control of clustered genotype data. By calculating “Silhouette scores”, the SNP genotype assignment can be evaluated by a single numeric measure. Silhouette scores were then applied in Study I to compare the performance of four DNA polymerases and in Study III to evaluate freeze-dried reagents in the tag-array minisequencing system. The characteristics of the tag-array minisequencing system makes it suitable for inexpensive genome-wide gene mapping in the fruit fly. In Study II a high-resolution SNP map, and 293 genotyping assays, were established across the X, 2nd and 3rd chromosomes to distinguish commonly used Drosophila strains. A database of the SNP markers and a program for automatic allele calling and identification of map positions of mutants was also developed. The utility of the system was demonstrated by rapid mapping of 14 genes that disrupt embryonic muscle patterning. In Study III the tag-array minisequencing system was adapted to a lab-on-a-chip format for diagnostic testing for mutations in the TP53 gene. Freeze-drying was evaluated for storing reagents, including thermo-sensitive enzymes, on the microchip to reduce the complexity of the integrated test. Correct genotyping results were obtained using freeze-dried reagents in each reaction step of the genotyping protocol, both in test tubes and in single polymer test chambers. The results showed the potential of the approach to be implemented in fully integrated systems. The four-colour chemistry of SBE has been developed further to allow massively parallel sequencing (MPS) of short DNA fragments as in the Genome Analyzer system (Solexa/Illumina). In Study IV MPS was used to compare Nimblegen arrays and the SureSelect solution-based system for targeted enrichment of 56 continuous human candidate-gene regions totalling 3.1 Mb in size. Both methods detected known SNPs and discovered novel SNPs in the target regions, demonstrating the feasibility for complexity reduction of sequencing libraries by hybridization methods.
|
|
| 2. |
|
|
| 3. |
- Axelsson, Morgan, et al.
(författare)
-
Column penetration tests for lime-cement columns in deep mixing - experiences in Sweden
- 2003
-
Ingår i: Geotechnical Special Publication. - 0895-0563. ; 120, s. 681-694
-
Tidskriftsartikel (refereegranskat)abstract
- In this paper column penetration tests for lime-cement columns in deep soil mixing are reviewed. In principle, column penetration tests and reversed column penetration tests are considered. Improved test techniques are presented and discussed based on tests from two test sites in Sweden. The investigations indicate that the reversed column penetration test is the most suitable method for the primary quality test with reference to the uniformity and continuity of the columns. The probe should, however, be installed by the lime-cement column machine short after the manufacturing of the column to avoid disturbances in the mixing process and to enable a random test selection.
|
|
| 4. |
- Azizpour, Hossein, 1985-, et al.
(författare)
-
From Generic to Specific Deep Representations for Visual Recognition
- 2015
-
Ingår i: Proceedings of CVPR 2015. - : IEEE conference proceedings. - 9781467367592
-
Konferensbidrag (refereegranskat)abstract
- Evidence is mounting that ConvNets are the best representation learning method for recognition. In the common scenario, a ConvNet is trained on a large labeled dataset and the feed-forward units activation, at a certain layer of the network, is used as a generic representation of an input image. Recent studies have shown this form of representation to be astoundingly effective for a wide range of recognition tasks. This paper thoroughly investigates the transferability of such representations w.r.t. several factors. It includes parameters for training the network such as its architecture and parameters of feature extraction. We further show that different visual recognition tasks can be categorically ordered based on their distance from the source task. We then show interesting results indicating a clear correlation between the performance of tasks and their distance from the source task conditioned on proposed factors. Furthermore, by optimizing these factors, we achieve stateof-the-art performances on 16 visual recognition tasks.
|
|
| 5. |
|
|
| 6. |
|
|
| 7. |
- Baldassarre, Federico, et al.
(författare)
-
Explainability Techniques for Graph Convolutional Networks
- 2019
-
Konferensbidrag (refereegranskat)abstract
- Graph Networks are used to make decisions in potentially complex scenarios but it is usually not obvious how or why they made them. In this work, we study the explainability of Graph Network decisions using two main classes of techniques, gradient-based and decomposition-based, on a toy dataset and a chemistry task. Our study sets the ground for future development as well as application to real-world problems.
|
|
| 8. |
- Baldassarre, Federico, et al.
(författare)
-
Explanation-Based Weakly-Supervised Learning of Visual Relations with Graph Networks
- 2020
-
Ingår i: Proceedings, Part XXVIII Computer Vision - ECCV 2020 - 16th European Conference, Glasgow, UK, August 23-28, 2020. - Cham : Springer Nature. ; , s. 612-630
-
Konferensbidrag (refereegranskat)abstract
- Visual relationship detection is fundamental for holistic image understanding. However, the localization and classification of (subject, predicate, object) triplets remain challenging tasks, due to the combinatorial explosion of possible relationships, their long-tailed distribution in natural images, and an expensive annotation process. This paper introduces a novel weakly-supervised method for visual relationship detection that relies on minimal image-level predicate labels. A graph neural network is trained to classify predicates in images from a graph representation of detected objects, implicitly encoding an inductive bias for pairwise relations. We then frame relationship detection as the explanation of such a predicate classifier, i.e. we obtain a complete relation by recovering the subject and object of a predicted predicate. We present results comparable to recent fully- and weakly-supervised methods on three diverse and challenging datasets: HICO-DET for human-object interaction, Visual Relationship Detection for generic object-to-object relations, and UnRel for unusual triplets; demonstrating robustness to non-comprehensive annotations and good few-shot generalization.
|
|
| 9. |
- Baldassarre, Federico
(författare)
-
Structured Representations for Explainable Deep Learning
- 2023
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Deep learning has revolutionized scientific research and is being used to take decisions in increasingly complex scenarios. With growing power comes a growing demand for transparency and interpretability. The field of Explainable AI aims to provide explanations for the predictions of AI systems. The state of the art of AI explainability, however, is far from satisfactory. For example, in Computer Vision, the most prominent post-hoc explanation methods produce pixel-wise heatmaps over the input domain, which are meant to visualize the importance of individual pixels of an image or video. We argue that such dense attribution maps are poorly interpretable to non-expert users because of the domain in which explanations are formed - we may recognize shapes in a heatmap but they are just blobs of pixels. In fact, the input domain is closer to the raw data of digital cameras than to the interpretable structures that humans use to communicate, e.g. objects or concepts. In this thesis, we propose to move beyond dense feature attributions by adopting structured internal representations as a more interpretable explanation domain. Conceptually, our approach splits a Deep Learning model in two: the perception step that takes as input dense representations and the reasoning step that learns to perform the task at hand. At the interface between the two are structured representations that correspond to well-defined objects, entities, and concepts. These representations serve as the interpretable domain for explaining the predictions of the model, allowing us to move towards more meaningful and informative explanations. The proposed approach introduces several challenges, such as how to obtain structured representations, how to use them for downstream tasks, and how to evaluate the resulting explanations. The works included in this thesis address these questions, validating the approach and providing concrete contributions to the field. For the perception step, we investigate how to obtain structured representations from dense representations, whether by manually designing them using domain knowledge or by learning them from data without supervision. For the reasoning step, we investigate how to use structured representations for downstream tasks, from Biology to Computer Vision, and how to evaluate the learned representations. For the explanation step, we investigate how to explain the predictions of models that operate in a structured domain, and how to evaluate the resulting explanations. Overall, we hope that this work inspires further research in Explainable AI and helps bridge the gap between high-performing Deep Learning models and the need for transparency and interpretability in real-world applications.
|
|
| 10. |
- Bassan, Arianna, et al.
(författare)
-
A Density Functional Study of a Biomimetic Non-Heme Iron Catalyst : Insights into Alkane Hydroxylation and Olefin Oxidation by a Formally HO-Fe(V)=O Oxidant
- 2004
-
Ingår i: Chemistry - A European Journal. - : Wiley. - 0947-6539 .- 1521-3765. ; 11:2, s. 692-705
-
Tidskriftsartikel (refereegranskat)abstract
- The reactivity of [HO(tpa)FeVO] (TPA=tris(2-pyridylmethyl)amine), derived from OO bond heterolysis of its [H2O(tpa)FeIIIOOH] precursor, was explored by means of hybrid density functional theory. The mechanism for alkane hydroxylation by the high-valent iron–oxo species invoked as an intermediate in Fe(tpa)/H2O2 catalysis was investigated. Hydroxylation of methane and propane by HOFeVO was studied by following the rebound mechanism associated with the heme center of cytochrome P450, and it is demonstrated that this species is capable of stereospecific alkane hydroxylation. The mechanism proposed for alkane hydroxylation by HOFeVO accounts for the experimentally observed incorporation of solvent water into the products. An investigation of the possible hydroxylation of acetonitrile (i.e., the solvent used in the experiments) shows that the activation energy for hydrogen-atom abstraction by HOFeVO is rather high and, in fact, rather similar to that of methane, despite the similarity of the HCH2CN bond strength to that of the secondary CH bond in propane. This result indicates that the kinetics of hydrogen-atom abstraction are strongly affected by the cyano group and rationalizes the lack of experimental evidence for solvent hydroxylation in competition with that of substrates such as cyclohexane.
|
|
