Library-based Approaches Research Articles

Transfer Learning and Few-Shot Learning Based Deep Neural Network Models for Underwater Sonar Image Classification With a Few Samples

Acoustic imaging sonar systems are widely used for long-range underwater surveillance in various civilian and military applications. They provide 2-D images of underwater objects, even in turbid water conditions where optical underwater imaging systems fail. Achieving high accuracy in automatic deep learning based underwater image classification remains an open problem due to insufficient data availability, poor image resolution, low signal-to-noise ratio surroundings, etc. In this study, we conduct a comparative analysis of different advanced deep learning approaches, i.e., transfer learning and few-shot learning, to address the problem of automatic object classification in sonar images, using a few samples of data. Specifically, two metric learning-based approaches, i.e., siamese network and triplet network as well as library-based approaches, are studied under the few-shot learning paradigm. Extensive experiments are conducted on a novel custom-made dataset developed in-house, along with the publicly available SeabedObjectsKLSG dataset. In addition, the effectiveness of the sampling technique in handling class imbalance during model training is also investigated in this work. Our experimental results highlight that the few-shot learning based approach is a promising direction for future research on underwater image classification with a few samples.

Super-Resolution without Imaging: Library-Based Approaches Using Near-to-Far-Field Transduction by a Nanophotonic Structure.

Super-resolution imaging is often viewed in terms of engineering narrow point spread functions, but nanoscale optical metrology can be performed without real-space imaging altogether. In this paper, we investigate how partial knowledge of scattering nanostructures enables extraction of nanoscale spatial information from far-field radiation patterns. We use principal component analysis to find patterns in calibration data and use these patterns to retrieve the position of a point source of light. In an experimental realization using angle-resolved cathodoluminescence, we retrieve the light source position with an average error below λ/100. The patterns found by principal component analysis reflect the underlying scattering physics and reveal the role the scattering nanostructure plays in localization success. The technique described here is highly general and can be applied to gain insight into and perform subdiffractive parameter retrieval in various applications.

Parallel Factor Analysis Enables Quantification and Identification of Highly Convolved Data-Independent-Acquired Protein Spectra

SummaryHigh-throughput data-independent acquisition (DIA) is the method of choice for quantitative proteomics, combining the best practices of targeted and shotgun approaches. The resultant DIA spectra are, however, highly convolved and with no direct precursor-fragment correspondence, complicating biological sample analysis. Here, we present CANDIA (canonical decomposition of data-independent-acquired spectra), a GPU-powered unsupervised multiway factor analysis framework that deconvolves multispectral scans to individual analyte spectra, chromatographic profiles, and sample abundances, using parallel factor analysis. The deconvolved spectra can be annotated with traditional database search engines or used as high-quality input for de novo sequencing methods. We demonstrate that spectral libraries generated with CANDIA substantially reduce the false discovery rate underlying the validation of spectral quantification. CANDIA covers up to 33 times more total ion current than library-based approaches, which typically use less than 5% of total recorded ions, thus allowing quantification and identification of signals from unexplored DIA spectra.

Coevolution of Adeno-associated Virus Capsid Antigenicity and Tropism through a Structure-Guided Approach.

Adeno-associated viruses (AAV) are composed of nonenveloped, icosahedral protein shells that can be adapted to package and deliver recombinant therapeutic DNA. Approaches to engineer recombinant capsids for gene therapy applications have focused on rational design or library-based approaches that can address one or two desirable attributes; however, there is an unmet need to comprehensively improve AAV vector properties. Such cannot be achieved by utilizing sequence data alone but requires harnessing the three-dimensional (3D) structural properties of AAV capsids. Here, we solve the structures of a natural AAV isolate complexed with antibodies using cryo-electron microscopy and harness this structural information to engineer AAV capsid libraries through saturation mutagenesis of different antigenic footprints. Each surface loop was evolved by infectious cycling in the presence of a helper adenovirus to yield a new AAV variant that then serves as a template for evolving the next surface loop. This stepwise process yielded a humanized AAV8 capsid (AAVhum.8) displaying nonnatural surface loops that simultaneously display tropism for human hepatocytes, increased gene transfer efficiency, and neutralizing antibody evasion. Specifically, AAVhum.8 can better evade neutralizing antisera from multiple species than AAV8. Further, AAVhum.8 displays robust transduction in a human liver xenograft mouse model with expanded tropism for both murine and human hepatocytes. This work supports the hypothesis that critical properties, such as AAV capsid antibody evasion and tropism, can be coevolved by combining rational design and library-based evolution for clinical gene therapy.IMPORTANCE Clinical gene therapy with recombinant AAV vectors has largely relied on natural capsid isolates. There is an unmet need to comprehensively improve AAV tissue tropism, transduction efficiency, and antibody evasion. Such cannot be achieved by utilizing capsid sequence data alone but requires harnessing the 3D structural properties of AAV capsids. Here, we combine rational design and library-based evolution to coevolve multiple, desirable properties onto AAV by harnessing 3D structural information.

Abstract 2183: Targeting TRBC1 and 2 for the treatment of T cell lymphomas

Abstract Background: Mature T cell lymphomas are aggressive, treatment resistant cancers that are associated with poor prognosis. Clinical application of immunotherapeutic approaches has been limited by a lack of target antigens that discriminate malignant from healthy T cells. Unlike B cell depletion, pan T cell aplasia is prohibitively toxic. The mutually exclusive expression of TCR β chains (TRBC) 1 or 2 allows targeting the malignant T cell population while preserving T cell function. Notably, a two amino acid inversion is the only exposed and antibody-accessible feature differentiating the two isoforms, making the development of therapeutics targeting TRBC1 and TRBC2 challenging. We have previously described a TRBC1 specific antibody which has been incorporated into a Chimeric Antigen Receptor (CAR) (Maciocia et al, Nat Med, 2017). Here we describe the derivation and characterization of a TRBC2 specific binder and CAR. In addition, we investigate use of these selective binders as Antibody Drug Conjugates (ADCs) for the treatment of HTLV-1 associated leukemia/lymphoma for which CAR T cell therapy may not be well suited. Methods: Anti TRBC2 antibodies were derived through crystallography and structural engineering of the previously described TRBC1 specific antibody. Briefly, TRBC2 binders were obtained by screening of small libraries of the TRBC1 binder with randomizations in key residues identified by crystallographic data and in silico design. TRBC2 binder candidates were first optimized for CAR function. Subsequently, TRBC2 specific CAR T cells were evaluated in vitro and in vivo for anti-tumor activity and selectivity. TRBC1/TRBC2 targeting antibodies, were further characterized as ADCs for biophysical properties, antibody internalization and cytotoxic function. Results: In vitro testing of TRBC2 CARs showed comparable efficacy to the previously described TRBC1 CAR. TRBC2 CAR T cells were effective at killing TRBC2 target cells while sparing TRBC1 positive cells. In vivo mouse models demonstrated that both TRBC1 and TRBC2 directed CARs could target their respective antigens with a high degree of specificity. TRBC1 and TRBC2 specific antibodies were used to generate ADC molecules as a proof of concept. Anti-TRBC1/TRBC2 antibodies were internalized and showed potential as ADCs. Interestingly, we demonstrate that an optimal affinity window facilitates improved antibody uptake and have further engineered both our TRBC1 and TRBC2 antibodies to take advantage of this particularity. Conclusions: Following on from structural and library-based approaches to generate CAR T cells capable of specifically targeting TRBC2, we have further characterized TRBC2 specific CAR T cells in vitro and in vivo. We have shown that highly specific antibodies, engineered as targeting moieties for TRBC1 and TRBC2 CAR T cells, show promising characteristics for utility potentially also as ADCs, offering another modality through which this targeting paradigm can be exploited for the treatment of peripheral T cell lymphomas. Citation Format: Mathieu Ferrari, Vania Baldan, Priyanka Ghongane, Alex Nicholson, Reyisa Bughda, Zulaikha Akbar, Patrycja Wawrzyniecka, Paul Maciocia, Shaun Cordoba, Simon Thomas, Shimobi Onuoha, Martin Pule. Targeting TRBC1 and 2 for the treatment of T cell lymphomas [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2183.

Catalyst: Can Systems Chemistry Unravel the Mysteries of the Chemical Origins of Life?

Daniela Kroiss obtained her MSc degree in biotechnology at BOKU and is currently a PhD student in the biochemistry program at CUNY, where she works on a joint project between the Nanoscience Initiative and the University of Strathclyde. Gonen Ashkenasy is head of the Laboratory for Systems Chemistry at the Ben-Gurion University Chemistry Department. His research focuses on the design and analysis of synthetic replication networks. Adam Braunschweig earned his PhD in organic chemistry at the University of California and is an associate professor in the CUNY Nanoscience Initiative. His research focuses on the interface of organic chemistry with biology and material science. Tell Tuttle obtained his PhD in theoretical chemistry at Goteborg University and is a professor of theoretical chemistry at the University of Strathclyde. His research focuses on the directed discovery of peptide-based materials. Rein Ulijn gained his PhD in physical chemistry at the University of Strathclyde and is founding director of the Nanoscience Initiative at the CUNY Advanced Science Research Center. His research focuses on nanotechnology inspired by living systems.

Residues on Adeno-associated Virus Capsid Lumen Dictate Interactions and Compatibility with the Assembly-Activating Protein.

The adeno-associated virus (AAV) serves as a broadly used vector system for in vivo gene delivery. The process of AAV capsid assembly remains poorly understood. The viral cofactor assembly-activating protein (AAP) is required for maximum AAV production and has multiple roles in capsid assembly, namely, trafficking of the structural proteins (VP) to the nuclear site of assembly, promoting the stability of VP against multiple degradation pathways, and facilitating stable interactions between VP monomers. The N-terminal 60 amino acids of AAP (AAPN) are essential for these functions. Presumably, AAP must physically interact with VP to execute its multiple functions, but the molecular nature of the AAP-VP interaction is not well understood. Here, we query how structurally related AAVs functionally engage AAP from AAV serotype 2 (AAP2) toward virion assembly. These studies led to the identification of key residues on the lumenal capsid surface that are important for AAP-VP and for VP-VP interactions. Replacing a cluster of glutamic acid residues with a glutamine-rich motif on the conserved VP beta-barrel structure of variants incompatible with AAP2 creates a gain-of-function mutant compatible with AAP2. Conversely, mutating positively charged residues within the hydrophobic region of AAP2 and conserved core domains within AAPN creates a gain-of-function AAP2 mutant that rescues assembly of the incompatible variant. Our results suggest a model for capsid assembly where surface charge/neutrality dictates an interaction between AAPN and the lumenal VP surface to nucleate capsid assembly.IMPORTANCE Efforts to engineer the AAV capsid to gain desirable properties for gene therapy (e.g., tropism, reduced immunogenicity, and higher potency) require that capsid modifications do not affect particle assembly. The relationship between VP and the cofactor that facilitates its assembly, AAP, is central to both assembly preservation and vector production. Understanding the requirements for this compatibility can inform manufacturing strategies to maximize production and reduce costs. Additionally, library-based approaches that simultaneously examine a large number of capsid variants would benefit from a universally functional AAP, which could hedge against overlooking variants with potentially valuable phenotypes that were lost during vector library production due to incompatibility with the cognate AAP. Studying interactions between the structural and nonstructural components of AAV enhances our fundamental knowledge of capsid assembly mechanisms and the protein-protein interactions required for productive assembly of the icosahedral capsid.

Motion Correction in Proton Resonance Frequency-based Thermometry in the Liver.

The unique ability of magnetic resonance imaging to measure temperature noninvasively, in vivo, makes it an attractive tool for monitoring interventional procedures, such as radiofrequency or microwave ablation in real-time. The most frequently used approach for magnetic resonance-based temperature measurement is proton resonance frequency (PRF) thermometry. Although it has many advantages, including tissue-independence and real-time capability, the main drawback is its motion sensitivity. This is likely the reason PRF thermometry in moving organs, such as the liver, is not commonly used in the clinical arena. In recent years, however, several developments suggest that motion-corrected thermometry in the liver is achievable. The present article summarizes the diverse attempts to correct thermometry in the liver. Therefore, the physical principle of PRF is introduced, with additional references for necrosis zone estimation and how to deal with fat phase modulation, and main magnetic field drifts. The primary categories of motion correction are presented, including general methods for motion compensation and library-based approaches, and referenceless thermometry and hybrid methods. Practical validation of the described methods in larger patient groups will be necessary to establish accurate motion-corrected thermometry in the clinical arena, with the goal of complete liver tumor ablation.

De novo design of signal sequences to localize cargo to the 1,2-propanediol utilization microcompartment.

Organizing heterologous biosyntheses inside bacterial cells can alleviate common problems owing to toxicity, poor kinetic performance, and cofactor imbalances. A subcellular organelle known as a bacterial microcompartment, such as the 1,2-propanediol utilization microcompartment of Salmonella, is a promising chassis for this strategy. Here we demonstrate de novo design of the N-terminal signal sequences used to direct cargo to these microcompartment organelles. We expand the native repertoire of signal sequences using rational and library-based approaches and show that a canonical leucine-zipper motif can function as a signal sequence for microcompartment localization. Our strategy can be applied to generate new signal sequences localizing arbitrary cargo proteins to the 1,2-propanediol utilization microcompartments.

Performing selections under dynamic conditions for synthetic biology applications.

As the design of synthetic circuits and metabolic networks becomes more complex it is often difficult to know a priori which parameters and design choices will result in a desired phenotype. To counter this, rational design can be complemented by library-based approaches where diversity is introduced and then coupled with screening or selection methods. Here, we used a model of competitive growth to show that selection can rapidly identify library variants with near-optimal phenotypes. Many synthetic biology applications require phenotypes that balance multiple objectives, such as responding to more than one chemical signal. In addition, desired traits may be time-dependent, for example changing with the growth phase. By applying dynamic inputs to the selection, we show that it is possible to select for traits that satisfy multiple goals. Furthermore, we demonstrate that the underlying diversity in a library is heavily influenced by the initial circuit design. Overall, our findings argue that rational synthetic circuit design, coupled with diversity generation and dynamic selection are powerful tools for many synthetic biology applications.

Antibody engineering for increased potency, breadth and half-life.

Purpose of reviewThis review highlights recent developments in HIV-1 antibody engineering and discusses the effects of increased polyreactivity on serum half-lives of engineered antibodies.Recent findingsRecent studies have uncovered a wealth of information about the relationship between the sequences and efficacies of anti-HIV-1 antibodies through a combination of bioinformatics, structural characterization and in vivo studies. This knowledge has stimulated efforts to enhance antibody breadth and potency for therapeutic use. Although some engineered antibodies have shown increased polyreactivity and short half-lives, promising efforts are circumventing these problems.SummaryAntibodies are desirable as therapeutics due to their ability to recognize targets with both specificity and high affinity. Furthermore, the ability of antibodies to stimulate Fc-mediated effector functions can increase their utility. Thus, mAbs have become central to strategies for the treatment of various diseases. Using both targeted and library-based approaches, antibodies can be engineered to improve their therapeutic properties. This article will discuss recent antibody engineering efforts to improve the breadth and potency of anti-HIV-1 antibodies. The polyreactivity of engineered HIV-1 bNAbs and the effect on serum half-life will be explored along with strategies to overcome problems introduced by engineering antibodies. Finally, advances in creating bispecific anti-HIV-1 reagents are discussed.

Proteolytic Post-translational Modification of Proteins: Proteomic Tools and Methodology

Proteolytic processing is a ubiquitous and irreversible post-translational modification involving limited and highly specific hydrolysis of peptide and isopeptide bonds of a protein by a protease. Cleavage generates shorter protein chains displaying neo-N and -C termini, often with new or modified biological activities. Within the past decade, degradomics and terminomics have emerged as significant proteomics subfields dedicated to characterizing proteolysis products as well as natural protein N and C termini. Here we provide an overview of contemporary proteomics-based methods, including specific quantitation, data analysis, and curation considerations, and highlight exciting new and emerging applications within these fields enabling in vivo analysis of proteolytic events.

MALDI TOF MS profiling of bacteria at the strain level: A review

Since the advent of the use of matrix-assisted laser desorption/ionization (MALDI) time-of-flight mass spectrometry (TOF MS) as a tool for microbial characterization, efforts to increase the taxonomic resolution of the approach have been made. The rapidity and efficacy of the approach have suggested applications in counter-bioterrorism, prevention of food contamination, and monitoring the spread of antibiotic-resistant bacteria. Strain-level resolution has been reported with diverse bacteria, using library-based and bioinformatics-enabled approaches. Three types of characterization at the strain level have been reported: strain categorization, strain differentiation, and strain identification. Efforts to enhance the library-based approach have involved sample pre-treatment and data reduction strategies. Bioinformatics approaches have leveraged the ever-increasing amount of publicly available genomic and proteomic data to attain strain-level characterization. Bioinformatics-enabled strategies have facilitated strain characterization via intact biomarker identification, bottom-up, and top-down approaches. Rigorous quantitative and advanced statistical analyses have fostered success at the strain level with both approaches. Library-based approaches can be limited by effects of sample preparation and culture conditions on reproducibility, whereas bioinformatics-enabled approaches are typically limited to bacteria, for which genetic and/or proteomic data are available. Biological molecules other than proteins produced in strain-specific manners, including lipids and lipopeptides, might represent other avenues by which strain-level resolution might be attained. Immunological and lectin-based chemistries have shown promise to enhance sensitivity and specificity. Whereas the limits of the taxonomic resolution of MALDI TOF MS profiling of bacteria appears bacterium-specific, recent data suggest that these limits might not yet have been reached.

Proteome Serological Determination of Tumor-Associated Antigens in Melanoma

Proteome serology may complement expression library-based approaches as strategy utilizing the patients' immune responses for the identification pathogenesis factors and potential targets for therapy and markers for diagnosis. Melanoma is a relatively immunogenic tumor and antigens recognized by melanoma-specific T cells have been extensively studied. The specificities of antibody responses to this malignancy have been analyzed to some extent by molecular genetic but not proteomics approaches. We screened sera of 94 melanoma patients for anti-melanoma reactivity and detected seropositivity in two-thirds of the patients with 2–6 antigens per case detected by 1D and an average of 2.3 per case by 2D Western blot analysis. For identification, antigen spots in Western blots were aligned with proteins in 2-DE and analyzed by mass spectrometry. 18 antigens were identified, 17 of which for the first time for melanoma. One of these antigens, galectin-3, has been related to various oncogenic processes including metastasis formation and invasiveness. Similarly, enolase has been found deregulated in different cancers. With at least 2 of 18 identified proteins implicated in oncogenic processes, the work confirms the potential of proteome-based antigen discovery to identify pathologically relevant proteins.

Library Selection Approaches to Engineering Enhanced Retroviral and Lentiviral Vectors

Retroviral and lentiviral based gene delivery vectors have been used in numerous pre-clinical studies and clinical trials due to their advantages, including stable and prolonged expression of therapeutic transgenes and minimal immune responses against the vector. Despite such advantages, however, retroviral vectors also have several limitations for gene therapy applications. For example, they can suffer from a lack of efficient or targeted gene delivery to key cell types. In addition, retroviral vector stability can be compromised by their envelope proteins. This review briefly describes how such limitations have been overcome by recently developed library selection approaches that borrow a lesson from nature: the ability of evolution to generate biomolecules with novel function. These library selection approaches are based on the construction of retroviral libraries where the sequences encoding natural viral components are partially randomized using a variety of methods in order to generate diverse libraries that can be selected to create improved or novel functions. These high throughput, library-based approaches provide a strong complement to rational engineering of viral components for the rapid development of efficient and safe retroviral and lentiviral vector systems for gene therapy.

Platform Synthesis: A Useful Strategy for Rapid and Systematic Generation of Molecular Diversity

Emergence of library-based approaches have changed the way of developing new functional molecules in materials science and pharmaceutical science. Therefore, reliable methods for rapid and systematic generation of functional molecules are highly called for in this field. We herein describe our concept of "platform synthesis" as a useful strategy for generating molecular diversity. This simple yet powerful strategy realizes the synthesis of a number of interesting multifunctional molecules, such as multisubstituted olefins, in a programmable and diversity-oriented format. As well as applications to the synthesis of pharmaceutically important molecules, such as tamoxifen and CDP840, applications to materials science, which have led to the discovery of interesting fluorescent materials and properties, are also described.

Identification of peptide ligands for malignancy- and growth-regulating galectins using random phage-display and designed combinatorial peptide libraries

Members of the galectin family of endogenous lectins are involved in tumor growth regulation and in establishing characteristics of the malignant phenotype via protein–carbohydrate and protein–protein interactions. To identify peptide ligands with the potential to modulate these tumor-relevant interactions beneficially, complementary screening methods were employed, that is, both phage-display and a combinatorial pentapeptide library with the key YXY tripeptide core. Three representative prototype galectins were selected. The search for high-affinity ligands among phage-displayed random heptamers yielded enrichment after five selection cycles of the nonglycomimetic CQSPSARSC peptide in the case of the chicken homologue of galectin-1 but not the human protein, an indication for specificity. The most active glycomimetic from the combinatorial library of 5832 pentamers was WYKYW. Identification of peptide ligands for galectins with and without glycomimetic properties is thus possible. Our study documents the potential to combine the two library-based approaches for structural optimization of lead peptides.

Assessment of statistical methods used in library-based approaches to microbial source tracking

Several commonly used statistical methods for fingerprint identification in microbial source tracking (MST) were examined to assess the effectiveness of pattern-matching algorithms to correctly identify sources. Although numerous statistical methods have been employed for source identification, no widespread consensus exists as to which is most appropriate. A large-scale comparison of several MST methods, using identical fecal sources, presented a unique opportunity to assess the utility of several popular statistical methods. These included discriminant analysis, nearest neighbour analysis, maximum similarity and average similarity, along with several measures of distance or similarity. Threshold criteria for excluding uncertain or poorly matched isolates from final analysis were also examined for their ability to reduce false positives and increase prediction success. Six independent libraries used in the study were constructed from indicator bacteria isolated from fecal materials of humans, seagulls, cows and dogs. Three of these libraries were constructed using the rep-PCR technique and three relied on antibiotic resistance analysis (ARA). Five of the libraries were constructed using Escherichia coli and one using Enterococcus spp. (ARA). Overall, the outcome of this study suggests a high degree of variability across statistical methods. Despite large differences in correct classification rates among the statistical methods, no single statistical approach emerged as superior. Thresholds failed to consistently increase rates of correct classification and improvement was often associated with substantial effective sample size reduction. Recommendations are provided to aid in selecting appropriate analyses for these types of data.

Highly parallel structured adaptive mesh refinement using parallel language-based approaches

Adaptive Mesh Refinement (AMR) calculations carried out on structured meshes play an exceedingly important role in several areas of science and engineering. This is so not just because AMR techniques allow us to carry out calculations very efficiently but also because they model very precisely the multi-scale fashion in which nature itself works. Many AMR applications are also amongst the most computationally intensive calculations undertaken making it necessary to use parallel supercomputers for their solution. While class library-based approaches are being attempted for parallel AMR we point out here that recent advances in the Fortran 90/95 standard and the OpenMP standard now make it possible to carry out highly parallel AMR calculations using language-based approaches. The language-based approaches offer several advantages over library-based approaches, the two principal ones being portability across parallel platforms and the best possible utilization of Distributed Shared Memory (DSM) hardware on machines that have such hardware. They also free up the applications scientist from being constrained by the static features of a class library. The choice of Fortran also ensures maximal reuse of pre-existing Fortran 77 applications and full Fortran 77-based processing efficiency on each computational node. Our implementation of the ideas presented here in the first author's RIEMANN framework essentially permits any serial, uniform grid, stencil-based Fortran code to be turned into a parallel AMR code. In this paper we first describe our strategy for using Fortran 90 in an object-oriented fashion. This permits AMR applications to be expressed in terms of familiar abstractions that are natural to the process of solving AMR hierarchies. We then describe the OpenMP features that are useful for parallel processing of AMR hierarchies in a load balanced fashion on multiprocessors. The automatic, parallel regridding of AMR hierarchies is also described. We then present a very efficient load balancer and show how it is to be used for load balanced solution of AMR hierarchies. Our load balancer is extremely general and should also see use in other disciplines. We follow this up with the application of the parallel AMR techniques developed here to the solution of elliptic and hyperbolic problems. For our elliptic problem we choose parallel, self-adaptive multigrid as an example. For our hyperbolic problem we choose time-dependent MHD as an example. In either case illustrative information is given about the adaptive processing of these systems. We also provide detailed scalability studies for both the above-mentioned problems which show that our methods scale extremely well up to several hundreds of processors.

Using distributed objects to build the Stanford digital library Infobus

For digital libraries to thrive, the providers of information processing services must be able to evolve their systems autonomously. However, as the complexity of their offerings increases, software tools more sophisticated than existing Web facilities are needed. Distributed object technology may be the answer. The availability of high-volume, increasingly sophisticated information is making the need for metadata facilities more urgent. Traditional, library-based approaches break down when used in an advanced digital library. More modular mechanisms are needed, and the CORBA system is one approach. Digital libraries are affected at a deep technical level by the widely differing user traditions of Web users and library patrons. The challenge and opportunity of digital libraries will be the synthesis of these traditions. The authors set out to create a technical infrastructure to support the construction of digital libraries. In their view, a digital library comprises widely distributed resources that can be maintained autonomously by different organizations and will not require adherence to uniform interfaces.