Implications of topological imbalance for representation learning on biomedical knowledge graphs.

Adoption of recently developed methods from machine learning has given rise to creation of drug-discovery knowledge graphs (KGs) that utilize the interconnected nature of the domain. Graph-based modelling of the data, combined with KG embedding (KGE) methods, are promising as they provide a more intuitive representation and are suitable for inference tasks such as predicting missing links. One common application is to produce ranked lists of genes for a given disease, where the rank is based on the perceived likelihood of association between the gene and the disease. It is thus critical that these predictions are not only pertinent but also biologically meaningful. However, KGs can be biased either directly due to the underlying data sources that are integrated or due to modelling choices in the construction of the graph, one consequence of which is that certain entities can get topologically overrepresented. We demonstrate the effect of these inherent structural imbalances, resulting in densely connected entities being highly ranked no matter the context. We provide support for this observation across different datasets, models as well as predictive tasks. Further, we present various graph perturbation experiments which yield more support to the observation that KGE models can be more influenced by the frequency of entities rather than any biological information encoded within the relations. Our results highlight the importance of data modelling choices, and emphasizes the need for practitioners to be mindful of these issues when interpreting model outputs and during KG composition.

Improving Peptide-Spectrum Matching by Fragmentation Prediction Using Hidden Markov Models.

Tandem mass spectrometry has become the method of choice for high-throughput, quantitative analysis in proteomics. Peptide spectrum matching algorithms score the concordance between the experimental and the theoretical spectra of candidate peptides by evaluating the number (or proportion) of theoretically possible fragment ions observed in the experimental spectra without any discrimination. However, the assumption that each theoretical fragment is just as likely to be observed is inaccurate. On the contrary, MS2 spectra often have few dominant fragments. Using millions of MS/MS spectra we show that there is high reproducibility across different fragmentation spectra given the precursor peptide and charge state, implying that there is a pattern to fragmentation. To capture this pattern we propose a novel prediction algorithm based on hidden Markov models with an efficient training process. We investigated the performance of our interpolated-HMM model, trained on millions of MS2 spectra, and found that our model picks up meaningful patterns in peptide fragmentation. Second, looking at the variability of the prediction performance by varying the train/test data split, we observed that our model performs well independent of the specific peptides that are present in the training data. Furthermore, we propose that the real value of this model is as a preprocessing step in the peptide identification process. The model can discern fragment ions that are unlikely to be intense for a given candidate peptide rather than using the actual predicted intensities. As such, probabilistic measures of concordance between experimental and theoretical spectra will leverage better statistics.

Molecular Portrait of Breast-Cancer-Derived Cell Lines Reveals Poor Similarity with Tumors.

Breast-cancer-derived cell lines are an important sample source for cancer proteomics and can be classified on the basis of transcriptomic analysis into subgroups corresponding to the molecular subtypes observed in mammary tumors. This study describes a tridimensional fractionation method that allows high sequence coverage and proteome-wide estimation of protein expression levels. This workflow has been used to conduct an in-depth quantitative proteomic survey of five breast cancer cell lines matching all major cancer subgroups and shows that despite their different classification, these cell lines display a very high level of similarity. A proteome-wide comparison with the RNA levels observed in the same samples showed very little to no correlation. Finally, we demonstrate that the proteomes of in vitro models of breast cancer display surprisingly little overlap with those of clinical samples.

11ß-HSD1 inhibition in men mitigates prednisolone-induced adverse effects in a proof-of-concept randomised double-blind placebo-controlled trial.

Glucocorticoids prescribed to limit inflammation, have significant adverse effects. As 11ß-hydroxysteroid dehydrogenase type 1 (11ß-HSD1) regenerates active glucocorticoid, we investigated whether 11ß-HSD1 inhibition with AZD4017 could mitigate adverse glucocorticoid effects without compromising their anti-inflammatory actions. We conducted a proof-of-concept, randomized, double-blind, placebo-controlled study at Research Unit, Churchill Hospital, Oxford, UK (NCT03111810). 32 healthy male volunteers were randomized to AZD4017 or placebo, alongside prednisolone treatment. Although the primary endpoint of the study (change in glucose disposal during a two-step hyperinsulinemic, normoglycemic clamp) wasn't met, hepatic insulin sensitivity worsened in the placebo-treated but not in the AZD4017-treated group. Protective effects of AZD4017 on markers of lipid metabolism and bone turnover were observed. Night-time blood pressure was higher in the placebo-treated but not in the AZD4017-treated group. Urinary (5aTHF+THF)/THE ratio was lower in the AZD4017-treated but remained the same in the placebo-treated group. Most anti-inflammatory actions of prednisolone persisted with AZD4017 co-treatment. Four adverse events were reported with AZD4017 and no serious adverse events. Here we show that co-administration of AZD4017 with prednisolone in men is a potential strategy to limit adverse glucocorticoid effects.

Multimodel pathway enrichment methods for functional evaluation of expression regulation.

Functional analysis of quantitative expression data is becoming common practice within the proteomics and transcriptomics fields; however, a gold standard for this type of analysis has yet not emerged. To grasp the systemic changes in biological systems, efficient and robust methods are needed for data analysis following expression regulation experiments. We discuss several conceptual and practical challenges potentially hindering the emergence of such methods and present a novel method, called FEvER, that utilizes two enrichment models in parallel. We also present analysis of three disparate differential expression data sets using our method and compare our results to other established methods. With many useful features such as pathway hierarchy overview, we believe the FEvER method and its software implementation will provide a useful tool for peers in the field of proteomics. Furthermore, we show that the method is also applicable to other types of expression data.

Protein expression changes in ovarian cancer during the transition from benign to malignant.

Epithelial ovarian carcinoma has in general a poor prognosis since the vast majority of tumors are genomically unstable and clinically highly aggressive. This results in rapid progression of malignancy potential while still asymptomatic and thus in late diagnosis. It is therefore of critical importance to develop methods to diagnose epithelial ovarian carcinoma at its earliest developmental stage, that is, to differentiate between benign tissue and its early malignant transformed counterparts. Here we present a shotgun quantitative proteomic screen of benign and malignant epithelial ovarian tumors using iTRAQ technology with LC-MALDI-TOF/TOF and LC-ESI-QTOF MS/MS. Pathway analysis of the shotgun data pointed to the PI3K/Akt signaling pathway as a significant discriminatory pathway. Selected candidate proteins from the shotgun screen were further confirmed in 51 individual tissue samples of normal, benign, borderline or malignant origin using LC-MRM analysis. The MRM profile demonstrated significant differences between the four groups separating the normal tissue samples from all tumor groups as well as perfectly separating the benign and malignant tumors with a ROC-area of 1. This work demonstrates the utility of using a shotgun approach to filter out a signature of a few proteins only that discriminates between the different sample groups.

Homology-directed repair protects the replicating genome from metabolic assaults.

Homology-directed repair (HDR) safeguards DNA integrity under various forms of stress, but how HDR protects replicating genomes under extensive metabolic alterations remains unclear. Here, we report that besides stalling replication forks, inhibition of ribonucleotide reductase (RNR) triggers metabolic imbalance manifested by the accumulation of increased reactive oxygen species (ROS) in cell nuclei. This leads to a redox-sensitive activation of the ATM kinase followed by phosphorylation of the MRE11 nuclease, which in HDR-deficient settings degrades stalled replication forks. Intriguingly, nascent DNA degradation by the ROS-ATM-MRE11 cascade is also triggered by hypoxia, which elevates signaling-competent ROS and attenuates functional HDR without arresting replication forks. Under these conditions, MRE11 degrades daughter-strand DNA gaps, which accumulate behind active replisomes and attract error-prone DNA polymerases to escalate mutation rates. Thus, HDR safeguards replicating genomes against metabolic assaults by restraining mutagenic repair at aberrantly processed nascent DNA. These findings have implications for cancer evolution and tumor therapy.

In Vitro Evolution of Antibodies Inspired by In Vivo Evolution.

In vitro generation of antibodies often requires variable domain sequence evolution to adapt the protein in terms of affinity, specificity, or developability. Such antibodies, including those that are of interest for clinical development, may have their origins in a diversity of immunoglobulin germline genes. Others and we have previously shown that antibodies of different origins tend to evolve along different, preferred trajectories. Apart from substitutions within the complementary determining regions, evolution may also, in a germline gene-origin-defined manner, be focused to residues in the framework regions, and even to residues within the protein core, in many instances at a substantial distance from the antibody's antigen-binding site. Examples of such germline origin-defined patterns of evolution are described. We propose that germline gene-preferred substitution patterns offer attractive alternatives that should be considered in efforts to evolve antibodies intended for therapeutic use with respect to appropriate affinity, specificity, and product developability. We also hypothesize that such germline gene-origin-defined in vitro evolution hold potential to result in products with limited immunogenicity, as similarly evolved antibodies will be parts of conventional, in vivo-generated antibody responses and thus are likely to have been seen by the immune system in the past.

Data on haplotype-supported immunoglobulin germline gene inference.

Data that defines IGHV (immunoglobulin heavy chain variable) germline gene inference using sequences of IgM-encoding transcriptomes obtained by Illumina MiSeq sequencing technology are described. Such inference is used to establish personalized germline gene sets for in-depth antibody repertoire studies and to detect new antibody germline genes from widely available immunoglobulin-encoding transcriptome data sets. Specifically, the data has been used to validate (Parallel antibody germline gene and haplotype analyses support the validity of immunoglobulin germline gene inference and discovery (DOI: 10.1016/j.molimm.2017.03.012) (Kirik et al., 2017) [1]) the inference process. This was accomplished based on analysis of the inferred germline genes' association to the donors' different haplotypes as defined by their different, expressed IGHJ alleles and/or IGHD genes/alleles. The data is important for development of validated germline gene databases containing entries inferred from immunoglobulin-encoding transcriptome sequencing data sets, and for generation of valid, personalized antibody germline gene repertoires.

Parallel antibody germline gene and haplotype analyses support the validity of immunoglobulin germline gene inference and discovery.

Analysis of antibody repertoire development and specific antibody responses important for e.g. autoimmune conditions, allergy, and protection against disease is supported by high throughput sequencing and associated bioinformatics pipelines that describe the diversity of the encoded antibody variable domains. Proper assignment of sequences to germline genes are important for many such processes, for instance in the analysis of somatic hypermutation. Germline gene inference from antibody-encoding transcriptomes, by using tools such as TIgGER or IgDiscover, has a potential to enhance the quality of such analyses. These tools may also be used to identify germline genes not previously known. In this study, we exploited such software for germline gene inference and define aspects of analysis settings and pre-existing knowledge of germline genes that affect the outcome of gene inference. Furthermore, we demonstrate the capacity of IGHJ and IGHD haplotype inference, whenever subjects are heterozygous with respect to such genes, to lend support to IGHV gene inference in general, and to the identification of novel alleles presently not recognized by germline gene reference directories. We propose that such haplotype analysis shall, whenever possible, be used in future best practice to support the outcome of germline gene inference. IGHJ-directed haplotype inference was also used to identify haplotypes not expressing some IGHV germline genes. In particular, we identified a haplotype that did not express several major germline genes such as IGHV1-8, IGHV3-9, IGHV3-15, IGHV1-18, IGHV3-21, and IGHV3-23. We envisage that haplotype analysis will provide an efficient approach to identify subjects for further studies of the link between the available immunoglobulin repertoire and outcomes of immune responses.

Antibody Heavy Chain Variable Domains of Different Germline Gene Origins Diversify through Different Paths.

B cells produce antibodies, key effector molecules in health and disease. They mature their properties, including their affinity for antigen, through hypermutation events; processes that involve, e.g., base substitution, codon insertion and deletion, often in association with an isotype switch. Investigations of antibody evolution define modes whereby particular antibody responses are able to form, and such studies provide insight important for instance for development of efficient vaccines. Antibody evolution is also used in vitro for the design of antibodies with improved properties. To better understand the basic concepts of antibody evolution, we analyzed the mutational paths, both in terms of amino acid substitution and insertions and deletions, taken by antibodies of the IgG isotype. The analysis focused on the evolution of the heavy chain variable domain of sets of antibodies, each with an origin in 1 of 11 different germline genes representing six human heavy chain germline gene subgroups. Investigated genes were isolated from cells of human bone marrow, a major site of antibody production, and characterized by next-generation sequencing and an in-house bioinformatics pipeline. Apart from substitutions within the complementarity determining regions, multiple framework residues including those in protein cores were targets of extensive diversification. Diversity, both in terms of substitutions, and insertions and deletions, in antibodies is focused to different positions in the sequence in a germline gene-unique manner. Altogether, our findings create a framework for understanding patterns of evolution of antibodies from defined germline genes.

Discovery-based protein expression profiling identifies distinct subgroups and pathways in leiomyosarcomas.

UNLABELLED: Soft tissue sarcomas (STS) are malignant tumors of mesenchymal origin. A substantial portion of these tumors exhibits complex karyotypes and lack characterized chromosomal aberrations. Owing to such properties, both histopathologic and molecular classification of these tumors has been a significant challenge. This study examines the protein expression of a large number of human STS, including subtype heterogeneity, using two-dimensional gel proteomics. In addition, detailed proteome profiles of a subset of pleomorphic STS specimens using an in-depth mass-spectrometry approach identified subgroups within the leiomyosarcomas with distinct protein expression patterns. Pathways analysis indicates that key biologic nodes like apoptosis, cytoskeleton remodeling, and telomere regulation are differentially regulated among these subgroups. Finally, investigating the similarities between protein expression of leiomyosarcomas and undifferentiated pleomorphic sarcomas (UPS) revealed similar protein expression profiles for these tumors, in comparison with pleomorphic leiomyosarcomas. IMPLICATIONS: These results suggest that UPS tumors share a similar lineage as leiomyosarcomas and are likely to originate from different stages of differentiation from mesenchymal stem cells to smooth muscle cells.