Systematic Assessment of Deep Learning-Based Predictors of Fragmentation Intensity Profiles.

In recent years, several deep learning-based methods have been proposed for predicting peptide fragment intensities. This study aims to provide a comprehensive assessment of six such methods, namely Prosit, DeepMass:Prism, pDeep3, AlphaPeptDeep, Prosit Transformer, and the method proposed by Guan et al. To this end, we evaluated the accuracy of the predicted intensity profiles for close to 1.7 million precursors (including both tryptic and HLA peptides) corresponding to more than 18 million experimental spectra procured from 40 independent submissions to the PRIDE repository that were acquired for different species using a variety of instruments and different dissociation types/energies. Specifically, for each method, distributions of similarity (measured by Pearson's correlation and normalized angle) between the predicted and the corresponding experimental b and y fragment intensities were generated. These distributions were used to ascertain the prediction accuracy and rank the prediction methods for particular types of experimental conditions. The effect of variables like precursor charge, length, and collision energy on the prediction accuracy was also investigated. In addition to prediction accuracy, the methods were evaluated in terms of prediction speed. The systematic assessment of these six methods may help in choosing the right method for MS/MS spectra prediction for particular needs.

A Simple Method for Robust and Accurate Intrinsic Subtyping of Breast Cancer.

Motivation: The PAM50 signature/method is widely used for intrinsic subtyping of breast cancer samples. However, depending on the number and composition of the samples included in a cohort, the method may assign different subtypes to the same sample. This lack of robustness is mainly due to the fact that PAM50 subtracts a reference profile, which is computed using all samples in the cohort, from each sample before classification. In this paper we propose modifications to PAM50 to develop a simple and robust single-sample classifier, called MPAM50, for intrinsic subtyping of breast cancer. Like PAM50, the modified method uses a nearest centroid approach for classification, but the centroids are computed differently, and the distances to the centroids are determined using an alternative method. Additionally, MPAM50 uses unnormalized expression values for classification and does not subtract a reference profile from the samples. In other words, MPAM50 classifies each sample independently, and so avoids the previously mentioned robustness issue. Results: A training set was employed to find the new MPAM50 centroids. MPAM50 was then tested on 19 independent datasets (obtained using various expression profiling technologies) containing 9637 samples. Overall good agreement was observed between the PAM50- and MPAM50-assigned subtypes with a median accuracy of 0.792, which (we show) is comparable with the median concordance between various implementations of PAM50. Additionally, MPAM50- and PAM50-assigned intrinsic subtypes were found to agree comparably with the reported clinical subtypes. Also, survival analyses indicated that MPAM50 preserves the prognostic value of the intrinsic subtypes. These observations demonstrate that MPAM50 can replace PAM50 without loss of performance. On the other hand, MPAM50 was compared with 2 previously published single-sample classifiers, and with 3 alternative modified PAM50 approaches. The results indicated a superior performance by MPAM50. Conclusions: MPAM50 is a robust, simple, and accurate single-sample classifier of intrinsic subtypes of breast cancer.

An 8-Gene Signature for Classifying Major Subtypes of Non-Small-Cell Lung Cancer.

Motivation: The precise diagnosis of the major subtypes, lung adenocarcinoma and lung squamous cell carcinoma, of non-small-cell lung cancer is of practical importance as some treatments are subtype-specific. However, in some cases diagnosis via the commonly-used method, that is staining the specimen using immunohistochemical markers, may be challenging. Hence, having a computational method that complements the diagnosis is desirable. In this paper, we propose a gene signature for this purpose. Results: We developed an expression-based method that systematically suggests a huge set of candidate gene signatures and finds the best candidate. By applying this method to a training set, the optimal gene signature was found by considering close to 765 billion candidate signatures. The 8-gene signature found for classifying the 2 aforementioned subtypes comprises TP63, CALML3, KRT5, PKP1, TESC, SPINK1, C9orf152, and KRT7. The signature achieved a high overall prediction accuracy of 0.936 when tested using 34 independent gene expression datasets obtained using different technologies and comprising 2556 adenocarcinoma and 1630 squamous cell carcinoma samples. Additionally, the signature performed well in clinically challenging cases, that is poorly differentiated tumors and specimens obtained from biopsies. In comparison with 2 previously reported signatures, our signature performed better in terms of overall accuracy and especially accuracy of classifying lung squamous cell carcinoma. Conclusions: Our signature is easy to use and accurate regardless of the technology used to obtain the gene expression profiles. It performs well even in clinically challenging cases and thus can assist pathologists in diagnosis of the ambiguous cases.

Exploring induced pluripotency in human fibroblasts via construction, validation, and application of a gene regulatory network.

Reprogramming of somatic cells to induced pluripotent stem cells, by overexpressing certain factors referred to as the reprogramming factors, can revolutionize regenerative medicine. To provide a coherent description of induced pluripotency from the gene regulation perspective, we use 35 microarray datasets to construct a reprogramming gene regulatory network. Comprising 276 nodes and 4471 links, the resulting network is, to the best of our knowledge, the largest gene regulatory network constructed for human fibroblast reprogramming and it is the only one built using a large number of experimental datasets. To build the network, a model that relates the expression profiles of the initial (fibroblast) and final (induced pluripotent stem cell) states is proposed and the model parameters (link strengths) are fitted using the experimental data. Twenty nine additional experimental datasets are collectively used to test the model/network, and good agreement between experimental and predicted gene expression profiles is found. We show that the model in conjunction with the constructed network can make useful predictions. For example, we demonstrate that our approach can incorporate the effect of reprogramming factor stoichiometry and that its predictions are consistent with the experimentally observed trends in reprogramming efficiency when the stoichiometric ratios vary. Using our model/network, we also suggest new (not used in training of the model) candidate sets of reprogramming factors, many of which have already been experimentally verified. These results suggest our model/network can potentially be used in devising new recipes for induced pluripotency with higher efficiencies. Additionally, we classify the links of the network into three classes of different importance, prioritizing them for experimental verification. We show that many of the links in the top ranked class are experimentally known to be important in reprogramming. Finally, comparing with other methods, we show that using our model is advantageous.

Ultra-High-Frequency Reprogramming of Individual Long-Term Hematopoietic Stem Cells Yields Low Somatic Variant Induced Pluripotent Stem Cells.

Efficiency of reprogramming of human cells into induced pluripotent stem cells (iPSCs) has remained low. We report that individual adult human CD49f+ long-term hematopoietic stem cells (LT-HSCs) can be reprogrammed into iPSCs at close to 50% efficiency using Sendai virus transduction. This exquisite sensitivity to reprogramming is specific to LT-HSCs, since it progressively decreases in committed progenitors. LT-HSC reprogramming can follow multiple paths and is most efficient when transduction is performed after the cells have exited G0. Sequencing of 75 paired skin fibroblasts/LT-HSC samples collected from nine individuals revealed that LT-HSCs contain a lower load of somatic single-nucleotide variants (SNVs) and indels than skin fibroblasts and accumulate about 12 SNVs/year. Mutation analysis revealed that LT-HSCs and fibroblasts have very different somatic mutation signatures and that somatic mutations in iPSCs generally exist prior to reprogramming. LT-HSCs may become the preferred cell source for the production of clinical-grade iPSCs.

Mechanism-based disease similarity.

In recent years several methods have been proposed to assign pairwise mechanism- based similarity scores to human diseases. Despite their differences in approach and performance, these methods work in a somewhat similar manner: first a set of biomolecules (genes, proteins, chemicals, etc.) is associated with each disease, and then a measure is defined to calculate the similarity between the sets assigned to a pair of diseases. Since the similarity score between two diseases is defined based on the underlying molecular processes, a high score may hint at a shared cause, and therefore a similar treatment, for both diseases. This is of great practical importance especially when a rare or newly-discovered disease, for which limited information is available, is found to be related to a disease with a known treatment. Thus, in this mini-review we briefly discuss the recently developed methods for computing mechanism-based disease- disease similarities.

Analytical solution and scaling of fluctuations in complex networks traversed by damped, interacting random walkers.

A general model for random walks (RWs) on networks is proposed. It incorporates damping and time-dependent links, and it includes standard (undamped, noninteracting) RWs (SRWs), coalescing RWs, and coalescing-branching RWs as special cases. The exact, time-dependent solutions for the average numbers of visits (w) to nodes and their fluctuations (σ2) are given, and the long-term σ-w relation is studied. Although σ ∝ w(1/2) for SRWs, this power law can be fragile when coalescing-branching interaction is present. Damping, however, often strengthens it but with an exponent generally different from 1/2.

DeCoaD: determining correlations among diseases using protein interaction networks.

BACKGROUND: Disease-disease similarities can be investigated from multiple perspectives. Identifying similar diseases based on the underlying biomolecular interactions can be especially useful, because it may shed light on the common causes of the diseases and therefore may provide clues for possible treatments. Here we introduce DeCoaD, a web-based program that uses a novel method to assign pair-wise similarity scores, called correlations, to genetic diseases. FINDINGS: DeCoaD uses a random walk to model the flow of information in a network within which nodes are either diseases or proteins and links signify either protein-protein interactions or disease-protein associations. For each protein node, the total number of visits by the random walker is called the weight of that node. Using a disease as both the starting and the terminating points of the random walks, a corresponding vector, whose elements are the weights associated with the proteins, can be constructed. The similarity between two diseases is defined as the cosine of the angle between their associated vectors. For a user-specified disease, DeCoaD outputs a list of similar diseases (with their corresponding correlations), and a graphical representation of the disease families that they belong to. Based on a probabilistic clustering algorithm, DeCoaD also outputs the clusters that the disease of interest is a member of, and the corresponding probabilities. The program also provides an interface to run enrichment analysis for the given disease or for any of the clusters that contains it. CONCLUSIONS: DeCoaD uses a novel algorithm to suggest non-trivial similarities between diseases with known gene associations, and also clusters the diseases based on their similarity scores. DeCoaD is available at http://www.ncbi.nlm.nih.gov/CBBresearch/Yu/mn/DeCoaD/.

Relating diseases by integrating gene associations and information flow through protein interaction network.

Identifying similar diseases could potentially provide deeper understanding of their underlying causes, and may even hint at possible treatments. For this purpose, it is necessary to have a similarity measure that reflects the underpinning molecular interactions and biological pathways. We have thus devised a network-based measure that can partially fulfill this goal. Our method assigns weights to all proteins (and consequently their encoding genes) by using information flow from a disease to the protein interaction network and back. Similarity between two diseases is then defined as the cosine of the angle between their corresponding weight vectors. The proposed method also provides a way to suggest disease-pathway associations by using the weights assigned to the genes to perform enrichment analysis for each disease. By calculating pairwise similarities between 2534 diseases, we show that our disease similarity measure is strongly correlated with the probability of finding the diseases in the same disease family and, more importantly, sharing biological pathways. We have also compared our results to those of MimMiner, a text-mining method that assigns pairwise similarity scores to diseases. We find the results of the two methods to be complementary. It is also shown that clustering diseases based on their similarities and performing enrichment analysis for the cluster centers significantly increases the term association rate, suggesting that the cluster centers are better representatives for biological pathways than the diseases themselves. This lends support to the view that our similarity measure is a good indicator of relatedness of biological processes involved in causing the diseases. Although not needed for understanding this paper, the raw results are available for download for further study at ftp://ftp.ncbi.nlm.nih.gov/pub/qmbpmn/DiseaseRelations/.

A method for the inclusion of sphenoidal electrodes in realistic EEG source imaging.

SUMMARY: Although EEG source imaging (ESI) has become more popular over the last few years, sphenoidal electrodes (SPE) have never been incorporated in ESI using realistic head models. This is in part because of the true locations of these electrodes are not exactly known. In this study, we demonstrate the feasibility of determining the true locations of SPE and incorporating this information into realistic ESI. The impact of including these electrodes in ESI in mesial temporal lobe epilepsy is also discussed. Seventeen patients were retrospectively selected for this study. To determine the positions of SPE in each case, two orthogonal x-rays (sagittal and coronal) of the SPE needle stilette were taken in the presence of previously digitized scalp electrodes. An in-house computer program was then used to find the locations of the tip of the needle stilette relative to the surface electrodes. These locations were then incorporated in a realistic head model based on the finite element method. EEG source imaging was then performed using averaged spikes for included patients suspected of having mesial temporal lobe epilepsy. Including SPE significantly shifted the ESI result even in the presence of subtemporal electrodes, resulting in an inferior and mesial displacement.

Automated removal of EKG artifact from EEG data using independent component analysis and continuous wavelet transformation.

The electrical potential produced by the cardiac activity sometimes contaminates electroencephalogram (EEG) recordings, resulting in spiky activities that are referred to as electrocardiographic (EKG) artifact. For a variety of reasons it is often desirable to automatically detect and remove these artifacts. Especially, for accurate source localization of epileptic spikes in an EEG recording from a patient with epilepsy, it is of great importance to remove any concurrent artifact. Due to similarities in morphology between the EKG artifacts and epileptic spikes, any automated artifact removal algorithm must have an extremely low false-positive rate in addition to a high detection rate. In this paper, an automated algorithm for removal of EKG artifact is proposed that satisfies such criteria. The proposed method, which uses combines independent component analysis and continuous wavelet transformation, uses both temporal and spatial characteristics of EKG related potentials to identify and remove the artifacts. The method outperforms algorithms that use general statistical features such as entropy and kurtosis for artifact rejection.

EEG source imaging in epilepsy--practicalities and pitfalls.

EEG source imaging (ESI) is a model-based imaging technique that integrates temporal and spatial components of EEG to identify the generating source of electrical potentials recorded on the scalp. Recent advances in computer technologies have made the analysis of ESI data less time-consuming, and have rekindled interest in this technique as a clinical diagnostic tool. On the basis of the available body of evidence, ESI seems to be a promising tool for epilepsy evaluation; however, the precise clinical value of ESI in presurgical evaluation of epilepsy and in localization of eloquent cortex remains to be investigated. In this Review, we describe two fundamental issues in ESI; namely, the forward and inverse problems, and their solutions. The clinical application of ESI in surgical planning for patients with medically refractory focal epilepsy, and its use in source reconstruction together with invasive recordings, is also discussed. As ESI can be used to map evoked responses, we discuss the clinical utility of this technique in cortical mapping-an essential process when planning resective surgery for brain regions that are in close proximity to eloquent cortex.

Sphenoidal electrodes significantly change the results of source localization of interictal spikes for a large percentage of patients with temporal lobe epilepsy.

Although scalp EEG is a very useful tool for presurgical evaluation in epilepsy, the 10-20 system of electrodes in many cases fails to accurately localize the source of the epileptic seizures. One suggested solution to this problem is to use additional electrodes. Sphenoidal electrodes especially have been suggested to be helpful in identifying the irritative and seizure onset zones in patients with temporal lobe epilepsy. However, the value of these electrodes has been debated, and in many epilepsy centers they are not used. In this study, we investigate the impact of sphenoidal electrodes by comparing the results of EEG source localization with and without sphenoidal recordings. We retrospectively selected patients with temporal lobe epilepsy based on their clinical semiology and electrophysiologic data. For each patient, a prototype spike was used as a template for an automatic pattern search to find similar activities. The identified spikes were then averaged and analyzed by fitting a dipole to the data. The recordings from sphenoidal electrodes were then excluded and the analysis was repeated. It was found that in more than half of the patients inclusion of sphenoidal electrodes resulted in a shift of more than 2 cm in the location of the fitted dipole, and in some cases moved the dipole from the frontal lobe or the insula to the temporal lobe. Our results suggest that sphenoidal electrodes are helpful in the analysis of the EEG recordings of patients suspected of having temporal lobe epilepsy.

A direct coupling between global and internal motions in a single domain protein? MD investigation of extreme scenarios.

Proteins are not rigid molecules, but exhibit internal motions on timescales ranging from femto- to milliseconds and beyond. In solution, proteins also experience global translational and rotational motions, sometimes on timescales comparable to those of the internal fluctuations. The possibility that internal and global motions may be directly coupled has intriguing implications, given that enzymes and cell signaling proteins typically associate with binding partners and cellular scaffolds. Such processes alter their global motion and may affect protein function. Here, we present molecular dynamics simulations of extreme case scenarios to examine whether a possible relationship exists. In our model protein, a ubiquitin-like RhoGTPase binding domain of plexin-B1, we removed either internal or global motions. Comparisons with unrestrained simulations show that internal and global motions are not appreciably coupled in this single-domain protein. This lack of coupling is consistent with the observation that the dynamics of water around the protein, which is thought to permit, if not stimulate, internal dynamics, is also largely independent of global motion. We discuss implications of these results for the structure and function of proteins.

Structure and function of the intracellular region of the plexin-b1 transmembrane receptor.

Members of the plexin family are unique transmembrane receptors in that they interact directly with Rho family small GTPases; moreover, they contain a GTPase-activating protein (GAP) domain for R-Ras, which is crucial for plexin-mediated regulation of cell motility. However, the functional role and structural basis of the interactions between the different intracellular domains of plexins remained unclear. Here we present the 2.4 A crystal structure of the complete intracellular region of human plexin-B1. The structure is monomeric and reveals that the GAP domain is folded into one structure from two segments, separated by the Rho GTPase binding domain (RBD). The RBD is not dimerized, as observed previously. Instead, binding of a conserved loop region appears to compete with dimerization and anchors the RBD to the GAP domain. Cell-based assays on mutant proteins confirm the functional importance of this coupling loop. Molecular modeling based on structural homology to p120(GAP).H-Ras suggests that Ras GTPases can bind to the plexin GAP region. Experimentally, we show that the monomeric intracellular plexin-B1 binds R-Ras but not H-Ras. These findings suggest that the monomeric form of the intracellular region is primed for GAP activity and extend a model for plexin activation.

Refinement of the primary hydration shell model for molecular dynamics simulations of large proteins.

A realistic representation of water molecules is important in molecular dynamics simulation of proteins. However, the standard method of solvating biomolecules, that is, immersing them in a box of water with periodic boundary conditions, is computationally expensive. The primary hydration shell (PHS) method, developed more than a decade ago and implemented in CHARMM, uses only a thin shell of water around the system of interest, and so greatly reduces the computational cost of simulations. Applying the PHS method, especially to larger proteins, revealed that further optimization and a partial reworking was required and here we present several improvements to its performance. The model is applied to systems with different sizes, and both water and protein behaviors are compared with those observed in standard simulations with periodic boundary conditions and, in some cases, with experimental data. The advantages of the modified PHS method over its original implementation are clearly apparent when it is applied to simulating the 82 kDa protein Malate Synthase G.

Insights into oncogenic mutations of plexin-B1 based on the solution structure of the Rho GTPase binding domain.

The plexin family of transmembrane receptors are important for axon guidance, angiogenesis, but also in cancer. Recently, plexin-B1 somatic missense mutations were found in both primary tumors and metastases of breast and prostate cancers, with several mutations mapping to the Rho GTPase binding domain (RBD) in the cytoplasmic region of the receptor. Here we present the NMR solution structure of this domain, confirming that the protein has both a ubiquitin-like fold and surface features. Oncogenic mutations T1795A and T1802A are located in a loop region, perturb the average structure locally, and have no effect on Rho GTPase binding affinity. Mutations L1815F and L1815P are located at the Rho GTPase binding site and are associated with a complete loss of binding for Rac1 and Rnd1. Both are found to disturb the conformation of the beta3-beta4 sheet and the orientation of surrounding side chains. Our study suggests that the oncogenic behavior of the mutants can be rationalized with reference to the structure of the RBD of plexin-B1.

Acceptable protein and solvent behavior in primary hydration shell simulations of hen lysozyme.

The "primary hydration shell" method in molecular dynamics simulations uses a two- to three-layer thick shell of explicitly represented water molecules as the solvent around the protein of interest. We show that despite its simplicity, this computationally cheap model is capable of predicting acceptable water and protein behavior using the CHARMM22/CMAP potential function. For protein dynamics, comparisons are made with Lipari-Szabo order parameters. These have been derived from NMR relaxation parameters for pico-nano second motions of the NH groups in the main-chain and NH(2) groups in Asn/Gln side chains in hen lysozyme. It is also shown that an even simpler, and therefore faster, water-shell model leads to results in similarly good agreement with experiments, and also compared with simulations using a full box of water with periodic boundary conditions or with an implicit solvation model. Thus, the primary hydration shell method should be useful in making larger systems accessible to extensive simulations.