Multiscale Laplacian Learning.

Machine learning has greatly influenced many fields, including science. However, despite of the tremendous accomplishments of machine learning, one of the key limitations of most existing machine learning approaches is their reliance on large labeled sets, and thus, data with limited labeled samples remains a challenge. Moreover, the performance of machine learning methods often severely hindered in case of diverse data, usually associated with smaller data sets or data associated with areas of study where the size of the data sets is constrained by high experimental cost and/or ethics. These challenges call for innovative strategies for dealing with these types of data. In this work, the aforementioned challenges are addressed by integrating graph-based frameworks, semi-supervised techniques, multiscale structures, and modified and adapted optimization procedures. This results in two innovative multiscale Laplacian learning (MLL) approaches for machine learning tasks, such as data classification, and for tackling data with limited samples, diverse data, and small data sets. The first approach, multikernel manifold learning (MML), integrates manifold learning with multikernel information and incorporates a warped kernel regularizer using multiscale graph Laplacians. The second approach, the multiscale MBO (MMBO) method, introduces multiscale Laplacians to the modification of the famous classical Merriman-Bence-Osher (MBO) scheme, and makes use of fast solvers. We demonstrate the performance of our algorithms experimentally on a variety of benchmark data sets, and compare them favorably to the state-of-art approaches.

From the reference human genome to human pangenome: Premise, promise and challenge.

The Reference Human Genome remains the single most important resource for mapping genetic variations and assessing their impact. However, it is monophasic, incomplete and not representative of the variation that exists in the population. Given the extent of ethno-geographic diversity and the consequent diversity in clinical manifestations of these variations, population specific references were developed overtime. The dramatically plummeting cost of sequencing whole genomes and the advent of third generation long range sequencers allowing accurate, error free, telomere-to-telomere assemblies of human genomes present us with a unique and unprecedented opportunity to develop a more composite standard reference consisting of a collection of multiple genomes that capture the maximal variation existing in the population, with the deepest annotation possible, enabling a realistic, reliable and actionable estimation of clinical significance of specific variations. The Human Pangenome Project thus is a logical next step promising a more accurate and global representation of genomic variations. The pangenome effort must be reciprocally complemented with precise variant discovery tools and exhaustive annotation to ensure unambiguous clinical assessment of the variant in ethno-geographical context. Here we discuss a broad roadmap, the challenges and way forward in developing a universal pangenome reference including data visualization techniques and integration of prior knowledge base in the new graph based architecture and tools to submit, compare, query, annotate and retrieve relevant information from the pangenomes. The biggest challenge, however, will be the ethical, legal and social implications and the training of human resource to the new reference paradigm.

[Graph-based and constraint-based heterologous metabolic pathway design methods and application].

It is among the goals in metabolic engineering to construct microbial cell factories producing high-yield and high value-added target products, and an important solution is to design efficient synthetic pathway for the target products. However, due to the difference in metabolic capacity among microbial chassises, the available substrate and the yielded products are limited. Therefore, it is urgent to design related metabolic pathways to improve the production capacity. Existing metabolic engineering approaches to designing heterologous pathways are mainly based on biological experience, which are inefficient. Moreover, the yielded results are in no way comprehensive. However, systems biology provides new methods for heterologous pathway design, particularly the graph-based and constraint-based methods. Based on the databases containing rich metabolism information, they search for and uncover possible metabolic pathways with designated strategy (graph-based method) or algorithm (constraint-based method) and then screen out the optimal pathway to guide the modification of strains. In this paper, we reviewed the databases and algorithms for pathway design, and the applications in metabolic engineering and discussed the strengths and weaknesses of existing algorithms in practical application, hoping to provide a reference for the selection of optimal methods for the design of product synthesis pathway.

Graph-based and constraint-based heterologous metabolic pathway design methods and application / 生物工程学报

It is among the goals in metabolic engineering to construct microbial cell factories producing high-yield and high value-added target products, and an important solution is to design efficient synthetic pathway for the target products. However, due to the difference in metabolic capacity among microbial chassises, the available substrate and the yielded products are limited. Therefore, it is urgent to design related metabolic pathways to improve the production capacity. Existing metabolic engineering approaches to designing heterologous pathways are mainly based on biological experience, which are inefficient. Moreover, the yielded results are in no way comprehensive. However, systems biology provides new methods for heterologous pathway design, particularly the graph-based and constraint-based methods. Based on the databases containing rich metabolism information, they search for and uncover possible metabolic pathways with designated strategy (graph-based method) or algorithm (constraint-based method) and then screen out the optimal pathway to guide the modification of strains. In this paper, we reviewed the databases and algorithms for pathway design, and the applications in metabolic engineering and discussed the strengths and weaknesses of existing algorithms in practical application, hoping to provide a reference for the selection of optimal methods for the design of product synthesis pathway.

Hyperbolic relational graph convolution networks plus: a simple but highly efficient QSAR-modeling method.

Accurate predictions of druggability and bioactivities of compounds are desirable to reduce the high cost and time of drug discovery. After more than five decades of continuing developments, quantitative structure-activity relationship (QSAR) methods have been established as indispensable tools that facilitate fast, reliable and affordable assessments of physicochemical and biological properties of compounds in drug-discovery programs. Currently, there are mainly two types of QSAR methods, descriptor-based methods and graph-based methods. The former is developed based on predefined molecular descriptors, whereas the latter is developed based on simple atomic and bond information. In this study, we presented a simple but highly efficient modeling method by combining molecular graphs and molecular descriptors as the input of a modified graph neural network, called hyperbolic relational graph convolution network plus (HRGCN+). The evaluation results show that HRGCN+ achieves state-of-the-art performance on 11 drug-discovery-related datasets. We also explored the impact of the addition of traditional molecular descriptors on the predictions of graph-based methods, and found that the addition of molecular descriptors can indeed boost the predictive power of graph-based methods. The results also highlight the strong anti-noise capability of our method. In addition, our method provides a way to interpret models at both the atom and descriptor levels, which can help medicinal chemists extract hidden information from complex datasets. We also offer an HRGCN+'s online prediction service at https://quantum.tencent.com/hrgcn/.