Machine learning approaches in MALDI-MSI: clinical applications

Introduction: Despite the unquestionable advantages of Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Imaging in visualizing the spatial distribution and the relative abundance of biomolecules directly on-tissue, the yielded data is complex and high dimensional. Therefore, analysis and interpretation of this huge amount of information is mathematically, statistically and computationally challenging.

Areas covered: This article reviews some of the challenges in data elaboration with particular emphasis on machine learning techniques employed in clinical applications, and can be useful in general as an entry point for those who want to study the computational aspects. Several characteristics of data processing are described, enlightening advantages and disadvantages. Different approaches for data elaboration focused on clinical applications are also provided. Practical tutorial based upon Orange Canvas and Weka software is included, helping familiarization with the data processing.

Expert commentary: Recently, MALDI-MSI has gained considerable attention and has been employed for research and diagnostic purposes, with successful results. Data dimensionality constitutes an important issue and statistical methods for information-preserving data reduction represent one of the most challenging aspects. The most common data reduction methods are characterized by collecting independent observations into a single table. However, the incorporation of relational information can improve the discriminatory capability of the data.     

Protein‐fold recognition using an improved single‐source K diverse shortest paths algorithm

Protein structure prediction, when construed as a fold recognition problem, is one of the most important applications of similarity search in bioinformatics. A new protein‐fold recognition method is reported which combines a single‐source K diverse shortest path (SSKDSP) algorithm with Enrichment of Network Topological Similarity (ENTS) algorithm to search a graphic feature space generated using sequence similarity and structural similarity metrics. A modified, more efficient SSKDSP algorithm is developed to improve the performance of graph searching. The new implementation of the SSKDSP algorithm empirically requires 82% less memory and 61% less time than the current implementation, allowing for the analysis of larger, denser graphs. Furthermore, the statistical significance of fold ranking generated from SSKDSP is assessed using ENTS. The reported ENTS‐SSKDSP algorithm outperforms original ENTS that uses random walk with restart for the graph search as well as other state‐of‐the‐art protein structure prediction algorithms HHSearch and Sparks‐X, as evaluated by a benchmark of 600 query proteins. The reported methods may easily be extended to other similarity search problems in bioinformatics and chemoinformatics. The SSKDSP software is available at http://compsci.hunter.cuny.edu/~leixie/sskdsp.html . Proteins 2016; 84:467–472. © 2016 Wiley Periodicals, Inc.     

基于蛋白互作网络识别非小细胞肺癌相关基因功能模块

本研究对非小细胞肺癌(non-small cell lung carcinoma,NSCLC)基因表达数据进行差异表达分析,并与蛋白质相互作用网络(PPIN)数据进行整合,进一步利用Heinz搜索算法识别NSCLC相关的基因功能模块,并对模块中的基因进行功能(GO term)和通路(KEGG)富集分析,旨在探究肺癌发病分子机制。蛋白互作网络分析得到一个包含96个基因和117个相互作用的功能模块,以及8个对NSCLC的发生和发展起到关键作用候选基因标志物。富集分析结果表明,这些基因主要富集于基因转录催化及染色质调控等生物学过程,并在基础转录因子、黏着连接、细胞周期、Wnt信号通路及HTLV-Ⅰ感染等生物学通路中发挥重要作用。本研究对非小细胞肺癌相关的基因和生物学通路进行预测,可用于肺癌的早期诊断和早期治疗,以降低肺癌死亡率。     

Dosimetric impact of statistical uncertainty on Monte Carlo dose calculation algorithm in volumetric modulated arc therapy using Monaco TPS for three different clinical cases

AimTo study the dosimetric impact of statistical uncertainty (SU) per plan on Monte Carlo (MC) calculation in Monaco? treatment planning system (TPS) during volumetric modulated arc therapy (VMAT) for three different clinical cases.BackgroundDuring MC calculation SU is an important factor to decide dose calculation accuracy and calculation time. It is necessary to evaluate optimal acceptance of SU for quality plan with reduced calculation time.Materials and methodsThree different clinical cases as the lung, larynx, and prostate treated using VMAT technique were chosen. Plans were generated with Monaco? V5.11 TPS with 2% statistical uncertainty. By keeping all other parameters constant, plans were recalculated by varying SU, 0.5%, 1%, 2%, 3%, 4%, and 5%. For plan evaluation, conformity index (CI), homogeneity index (HI), dose coverage to PTV, organ at risk (OAR) dose, normal tissue receiving dose ≥5 Gy and ≥10 Gy, integral dose (NTID), calculation time, gamma pass rate, calculation reproducibility and energy dependency were analyzed.ResultsCI and HI improve as SU increases from 0.5% to 5%. No significant dose difference was observed in dose coverage to PTV, OAR doses, normal tissue receiving dose ≥5 Gy and ≥10 Gy and NTID. Increase of SU showed decrease in calculation time, gamma pass rate and increase in PTV max dose. No dose difference was seen in calculation reproducibility and dependent on energy.ConclusionFor VMAT plans, SU can be accepted from 1% to 3% per plan with reduced calculation time without compromising plan quality and deliverability by accepting variations in point dose within the target.     

Assigning alleles to different loci in amplifications of duplicated loci

Duplicated loci, for example those associated with major histocompatibility complex (MHC) genes, often have similar DNA sequences that can be coamplified with a pair of primers. This results in genotyping difficulties and inaccurate analyses. Here, we present a method to assign alleles to different loci in amplifications of duplicated loci. This method simultaneously considers several factors that may each affect correct allele assignment. These are the sharing of identical alleles among loci, null alleles, copy number variation, negative amplification, heterozygote excess or heterozygote deficiency, and linkage disequilibrium. The possible multilocus genotypes are extracted from the alleles for each individual and weighted to estimate the allele frequencies. The likelihood of an allele configuration is calculated and is optimized with a heuristic algorithm. Monte‐Carlo simulations and three empirical MHC data sets are used as examples to evaluate the efficacy of our method under different conditions. Our new software, mhc‐typer V1.1, is freely available at https://github.com/huangkang1987/mhc-typer .     

Structural insights into the catalytic mechanism of 5-methylthioribose 1-phosphate isomerase

5-Methylthioribose 1-phosphate isomerase (M1Pi) is a crucial enzyme involved in the universally conserved methionine salvage pathway (MSP) where it is known to catalyze the conversion of 5-methylthioribose 1-phosphate (MTR-1-P) to 5-methylthioribulose 1-phosphate (MTRu-1-P) via a mechanism which remains unspecified till date. Furthermore, although M1Pi has a discrete function, it surprisingly shares high structural similarity with two functionally non-related proteins such as ribose-1,5-bisphosphate isomerase (R15Pi) and the regulatory subunits of eukaryotic translation initiation factor 2B (eIF2B). To identify the distinct structural features that lead to divergent functional obligations of M1Pi as well as to understand the mechanism of enzyme catalysis, the crystal structure of M1Pi from a hyperthermophilic archaeon Pyrococcus horikoshii OT3 was determined. A meticulous structural investigation of the dimeric M1Pi revealed the presence of an N-terminal extension and a hydrophobic patch absent in R15Pi and the regulatory α-subunit of eIF2B. Furthermore, unlike R15Pi in which a kink formation is observed in one of the helices, the domain movement of M1Pi is distinguished by a forward shift in a loop covering the active-site pocket. All these structural attributes contribute towards a hydrophobic microenvironment in the vicinity of the active site of the enzyme making it favorable for the reaction mechanism to commence. Thus, a hydrophobic active-site microenvironment in addition to the availability of optimal amino-acid residues surrounding the catalytic residues in M1Pi led us to propose its probable reaction mechanism via a cis-phosphoenolate intermediate formation.     

Walsh Hadamard Transform for Simple Linear Iterative Clustering (SLIC) Superpixel Based Spectral Clustering of Multimodal MRI Brain Tumor Segmentation

The automated brain tumor segmentation methods are challenging due to the diverse nature of tumors. Recently, the graph based spectral clustering method is utilized for brain tumor segmentation to make high-quality segmentation output. In this paper, a new Walsh Hadamard Transform (WHT) texture for superpixel based spectral clustering is proposed for segmentation of a brain tumor from multimodal MRI images. First, the selected kernels of WHT are utilized for creating texture saliency maps and it becomes the input for the Simple Linear Iterative Clustering (SLIC) algorithm, to generate more precise texture based superpixels. Then the texture superpixels become nodes in the graph of spectral clustering for segmenting brain tumors of MRI images. Finally, the original members of superpixels are recovered to represent Complete Tumor (CT), Tumor Core (TC) and Enhancing Tumor (ET) tissues. The observational results are taken out on BRATS 2015 datasets and evaluated using the Dice Score (DS), Hausdorff Distance (HD) and Volumetric Difference (VD) metrics. The proposed method produces competitive results than other existing clustering methods.     

Smart process development: Application of machine-learning and integrated process modeling for inclusion body purification processes

The development of a biopharmaceutical production process usually occurs sequentially, and tedious optimization of each individual unit operation is very time-consuming. Here, the conditions established as optimal for one-step serve as input for the following step. Yet, this strategy does not consider potential interactions between a priori distant process steps and therefore cannot guarantee for optimal overall process performance. To overcome these limitations, we established a smart approach to develop and utilize integrated process models using machine learning techniques and genetic algorithms. We evaluated the application of the data-driven models to explore potential efficiency increases and compared them to a conventional development approach for one of our development products. First, we developed a data-driven integrated process model using gradient boosting machines and Gaussian processes as machine learning techniques and a genetic algorithm as recommendation engine for two downstream unit operations, namely solubilization and refolding. Through projection of the results into our large-scale facility, we predicted a twofold increase in productivity. Second, we extended the model to a three-step model by including the capture chromatography. Here, depending on the selected baseline-process chosen for comparison, we obtained between 50% and 100% increase in productivity. These data show the successful application of machine learning techniques and optimization algorithms for downstream process development. Finally, our results highlight the importance of considering integrated process models for the whole process chain, including all unit operations.