Data infrastructures for AI in medical imaging: a report on the experiences of five EU projects

Artificial intelligence (AI) is transforming the field of medical imaging and has the potential to bring medicine from the era of 'sick-care' to the era of healthcare and prevention. The development of AI requires access to large, complete, and harmonized real-world datasets, representative of the population, and disease diversity. However, to date, efforts are fragmented, based on single-institution, size-limited, and annotation-limited datasets. Available public datasets (e.g., The Cancer Imaging Archive, TCIA, USA) are limited in scope, making model generalizability really difficult. In this direction, five European Union projects are currently working on the development of big data infrastructures that will enable European, ethically and General Data Protection Regulation-compliant, quality-controlled, cancer-related, medical imaging platforms, in which both large-scale data and AI algorithms will coexist. The vision is to create sustainable AI cloud-based platforms for the development, implementation, verification, and validation of trustable, usable, and reliable AI models for addressing specific unmet needs regarding cancer care provision. In this paper, we present an overview of the development efforts highlighting challenges and approaches selected providing valuable feedback to future attempts in the area.Key points• Artificial intelligence models for health imaging require access to large amounts of harmonized imaging data and metadata.• Main infrastructures adopted either collect centrally anonymized data or enable access to pseudonymized distributed data.• Developing a common data model for storing all relevant information is a challenge.• Trust of data providers in data sharing initiatives is essential.• An online European Union meta-tool-repository is a necessity minimizing effort duplication for the various projects in the area.

The Quest for Orthologs benchmark service and consensus calls in 2020

The identification of orthologs—genes in different species which descended from the same gene in their last common ancestor—is a prerequisite for many analyses in comparative genomics and molecular evolution. Numerous algorithms and resources have been conceived to address this problem, but benchmarking and interpreting them is fraught with difficulties (need to compare them on a common input dataset, absence of ground truth, computational cost of calling orthologs). To address this, the Quest for Orthologs consortium maintains a reference set of proteomes and provides a web server for continuous orthology benchmarking (http://orthology.benchmarkservice.org). Furthermore, consensus ortholog calls derived from public benchmark submissions are provided on the Alliance of Genome Resources website, the joint portal of NIH-funded model organism databases.

BioExcel Building Blocks, a software library for interoperable biomolecular simulation workflows

In the recent years, the improvement of software and hardware performance has made biomolecular simulations a mature tool for the study of biological processes. Simulation length and the size and complexity of the analyzed systems make simulations both complementary and compatible with other bioinformatics disciplines. However, the characteristics of the software packages used for simulation have prevented the adoption of the technologies accepted in other bioinformatics fields like automated deployment systems, workflow orchestration, or the use of software containers. We present here a comprehensive exercise to bring biomolecular simulations to the “bioinformatics way of working”. The exercise has led to the development of the BioExcel Building Blocks (BioBB) library. BioBB’s are built as Python wrappers to provide an interoperable architecture. BioBB’s have been integrated in a chain of usual software management tools to generate data ontologies, documentation, installation packages, software containers and ways of integration with workflow managers, that make them usable in most computational environments.

The carboxy-terminal region of human lipoprotein lipase is necessary for its exit from the endoplasmic reticulum

Certain missense substitutions on the human lipase (hLPL) gene produce mutated proteins that are retained in different compartments along the secretory pathway. The purpose of the present study was to elucidate whether the C-terminal domain of the hLPL molecule could be important for secretion. We constructed by site-directed mutagenesis three carboxy-terminal mutated (F388-->Stop, K428-->Stop and K441-->Stop) hLPL cDNAs that were expressed in COS1 cells. Immunoblotting of cell extracts showed that all three constructs led to similar levels of protein. Both wild type (WT) hLPL and the truncated K441-->Stop hLPL were secreted to the extracellular medium, and presented a similar intracellular distribution pattern as shown by immunofluorescence. Neither F388-->Stop nor K428-->Stop hLPL protein was detected in cell medium. Immunofluorescence experiments showed that both truncated hLPL were retained within an intracellular compartment, which became larger. Double immunofluorescence analysis using antibodies against LPL and antiprotein disulfide isomerase as a marker showed that the truncated K428-->Stop hLPL was retained within the rough endoplasmic reticulum. This truncated protein was not found in other compartments in the secretory pathway, such as Golgi complex and lysosomes, indicating that it did not exit the endoplasmic reticulum. Further analysis of the C-terminal region of the LPL molecular model showed both that F388-->Stop and K428-->Stop hLPL truncated proteins are highly hydrophobic. As retention of secretory proteins in the rough endoplasmic reticulum is a quality control mechanism of the secretory pathway, we conclude that the C-terminal domain of hLPL is critical for correct intracellular processing of the newly synthesized protein.

D-2-hydroxy-4-methylvalerate dehydrogenase from Lactobacillus delbrueckii subsp. bulgaricus. II. Mutagenic analysis of catalytically important residues

Five residues involved in catalysis and coenzyme binding have been identified in D-2-hydroxy-4-methylvalerate dehydrogenase from Lactobacillus delbrueckii subsp. bulgaricus by using biochemical and genetical methods. Enzyme inactivation with diethylpyrocarbonate indicated that a single histidine residue was involved in catalysis. Since H296 is the only conserved histidine in the whole family of NAD-dependent D-2-hydroxyacid dehydrogenases, we constructed the H296Q and H296S mutants and showed that their catalytic efficiencies were reduced 10(5)-fold compared with the wild-type enzyme. This low residual activity was shown to be insensitive to diethylpyrocarbonate. Taken together these data demonstrate that H296 is responsible for proton exchange in the redox reaction. Two acidic residues (D259 and E264) were candidates for maintaining H296 in the protonated state and their roles were examined by mutagenesis. The D259N and E264Q mutant enzymes both showed similar and large reductions in their Kcat/K(m) ratios (200-800-fold, depending on pH), indicating that either D259 or E264 (or both) could partner H296. The conserved R235 residue was a candidate for binding the alpha-carboxyl group of the substrate and it was changed to lysine. The R235K mutant showed a 104-fold reduced Kcat/K(m) due to both an increased K(m) and a reduced Kcat for 2-oxo-4-methylvalerate. Thus R235 plays a role in binding the substrate carboxylate similar to R171 in the L-lactate dehydrogenases. Finally, we constructed the H205Q mutant to test the role of this partially conserved histidine residue (in 10/13 enzymes of the family). This mutant enzyme displayed a 7.7-fold increased Kcat and a doubled catalytic efficiency at pH 5, was as sensitive to diethylpyrocarbonate as the wild-type but showed a sevenfold increased K(m) for NADH and a 100-fold increase in Kd for NADH together with 10-30-fold lower substrate inhibition. The transient kinetic behaviour of the H205Q mutant is as predicted from our previous study on the enzymatic mechanism of D-2-hydroxy-4-methylvalerate dehydrogenase which showed that coenzyme binding is highly pH dependent and indicated that release of the oxidised coenzyme is a significant component of the rate-limiting processes in catalysis at pH 6.5.

Contribution of a buried aspartate residue towards the catalytic efficiency and structural stability of Bacillus stearothermophilus lactate dehydrogenase

The X-ray structure of lactate dehydrogenase (LDH) shows the side-chain carboxylate group of Asp-143 to be buried in the hydrophobic interior of the enzyme, where it makes hydrogen-bonding interactions with both the side-chain hydroxyl group of Ser-273 and the main-chain amide group of His-195. This is an unusual environment for a carboxylate side-chain as hydrogen bonding normally occurs with water molecules at the surface of the protein. A charged hydrogen-bonding interaction in the interior of a protein would be expected to be much stronger than a similar interaction on the solvent-exposed exterior. In this respect the side-chain carboxylate group of Asp-143 appears to be important for maintaining tertiary structure by providing a common linkage point between three discontinuous elements of the secondary structure, alpha 1F, beta K and the beta-turn joining beta G and beta H. The contribution of the Asp-143 side-chain to the structure and function of Bacillus stearothermophilus LDH was assessed by creating a mutant enzyme containing Asn-143. The decreased thermal stability of both unactivated and fructose-1,6-diphosphate (Fru-1,6-P2)-activated forms of the mutant enzyme support a structural role for Asp-143. Furthermore, the difference in stability of the wild-type and mutant enzymes in guanidinium chloride suggested that the carboxylate group of Asp-143 contributes at least 22 kJ/mol to the conformational stability of the wild-type enzyme. However, there was no alteration in the amount of accessible tryptophan fluorescence in the mutant enzyme, indicating that the mutation caused a structural weakness rather than a gross conformational change. Comparison of the wild-type and mutant enzyme steady-state parameters for various 2-keto acid substrates showed the mutation to have a general effect on catalysis, with an average difference in binding energy of 11 kJ/mol for the transition-state complexes. The different effects of pH and Fru-1,6-P2 on the wild-type and mutant enzymes also confirmed a perturbation of the catalytic centre in the mutant enzyme. As the side-chain of Asp-143 is not sufficiently close to the active site to be directly involved in catalysis or substrate binding it is proposed that the effects on catalysis shown by the mutant enzyme are induced either by a structural change or by charge imbalance at the active site.(ABSTRACT TRUNCATED AT 400 WORDS)

Construction of a stable dimer of Bacillus stearothermophilus lactate dehydrogenase

A molecular graphics analysis of the features which prevent cytosolic malate dehydrogenase dimers from forming tetramers was evaluated by its success in predicting the synthesis of a version of the LDH framework which is a stable dimer. Surface residues responsible for malate dehydrogenases being dimers were revealed by superimposing the structures of two dimers of pig cytosolic malate dehydrogenase on one homologous tetramer of L-lactate dehydrogenase from Bacillus stearothermophilus. Four regions were identified as composing the P-axis dimer-dimer interface. Two regions of the dimer were surface loops that collided when built as a tetramer: a large loop (residues 203-207, KNOBI) and a small loop (residues 264-269, KNOBII), and these were candidates to explain the dimeric character of malate dehydrogenase. The analysis was tested by constructing a synthetic B. stearothermophilus lactate dehydrogenase (KNOBI) containing the large malate dehydrogenase loop (residues 203-207 being AYIKLQAKE, and extra four amino acids). The new construct was thermotolerant (90 degrees C) and enzymically active with kcat and KM (pyruvate) values similar to those of the wild-type enzyme. However, whereas the allosteric activator fructose 1,6-bisphosphate decreased KM 100 times for wild type, it had no influence on KNOBI. The molecular volumes of 1-120 microM concentrations of the construct were measured by time-resolved decay of tryptophan fluorescence anisotropy and by gel filtration. Both methods showed the molecular weight of wild type increased from dimer to tetramer with Kd about 20 microM dimer. KNOBI remained a dimer under these conditions.(ABSTRACT TRUNCATED AT 250 WORDS)