Small Data Can Play a Big Role in Chemical Discovery

The chemistry community is currently witnessing a surge of scientific discoveries in organic chemistry supported by machine learning (ML) techniques. Whereas many of these techniques were developed for big data applications, the nature of experimental organic chemistry often confines practitioners to small datasets. Herein, we touch upon the limitations associated with small data in ML and emphasize the impact of bias and variance on constructing reliable predictive models. We aim to raise awareness to these possible pitfalls, and thus, provide an introductory guideline for good practice. Ultimately, we stress the great value associated with statistical analysis of small data, which can be further boosted by adopting a holistic data-centric approach in chemistry.     

Synthetic Matching of Complex Monoterpene Indole Alkaloid Chemical Space

Monoterpene indole alkaloids (MIAs) are endowed with high structural and spatial complexity and characterized by diverse biological activities. Given this complexity-activity combination in MIAs, rapid and efficient access to chemical matter related to and with complexity similar to these alkaloids would be highly desirable, since such compound classes might display novel bioactivity. We describe the design and synthesis of a pseudo-natural product (pseudo-NP) collection obtained by the unprecedented combination of MIA fragments through complexity-generating transformations, resulting in arrangements not currently accessible by biosynthetic pathways. Cheminformatic analyses revealed that both the pseudo-NPs and the MIAs reside in a unique and common area of chemical space with high spatial complexity-density that is only sparsely populated by other natural products and drugs. Investigation of bioactivity guided by morphological profiling identified pseudo-NPs that inhibit DNA synthesis and modulate tubulin. These results demonstrate that the pseudo-NP collection occupies similar biologically relevant chemical space that Nature has endowed MIAs with.     

基于CouchDB和Elastic Search的高性能化学结构搜索引擎与数据库的构建

计算机辅助的化学结构搜索在化学信息学中地位十分重要,本文设计了一套高性能的化学结构和化学数据搜索系统,称为DCAIKU.DCAIKU基于CouchDB无模式数据库和ElasticSearch基础架构构建,通过将结构相似性搜索变换为文字搜索实现了高性能和高灵活性的检索引擎：在满足化学信息存储的高灵活性条件下,仍然可以做到低延迟和高准确性,同时拥有良好的伸缩性,可以大规模并行化和集群化.     

大学本科化学信息学课程的设计与优化

在分析了化学信息学课程基本内容与特点的基础之上,结合近年来我院相关课程的教学改革与实践,提出了大学化学信息学教学的两个层次;重点探讨了本科生化学信息学课程体系的设计与优化,为推进化学信息学教学改革抛砖引玉。     

Informatics for cross-sample analysis with comprehensive two-dimensional gas chromatography and high-resolution mass spectrometry (GCxGC-HRMS)

This paper describes informatics for cross-sample analysis with comprehensive two-dimensional gas chromatography (GCxGC) and high-resolution mass spectrometry (HRMS). GCxGC-HRMS analysis produces large data sets that are rich with information, but highly complex. The size of the data and volume of information requires automated processing for comprehensive cross-sample analysis, but the complexity poses a challenge for developing robust methods. The approach developed here analyzes GCxGC-HRMS data from multiple samples to extract a feature template that comprehensively captures the pattern of peaks detected in the retention-times plane. Then, for each sample chromatogram, the template is geometrically transformed to align with the detected peak pattern and generate a set of feature measurements for cross-sample analyses such as sample classification and biomarker discovery. The approach avoids the intractable problem of comprehensive peak matching by using a few reliable peaks for alignment and peak-based retention-plane windows to define comprehensive features that can be reliably matched for cross-sample analysis. The informatics are demonstrated with a set of 18 samples from breast-cancer tumors, each from different individuals, six each for Grades 1-3. The features allow classification that matches grading by a cancer pathologist with 78% success in leave-one-out cross-validation experiments. The HRMS signatures of the features of interest can be examined for determining elemental compositions and identifying compounds.     

Application of an integrated cheminformatics-molecular docking approach for discovery for physicochemically similar analogs of fluoroquinolones as putative HCV inhibitors

BackgroundHepatitis C Virus (HCV) infection is a major public health concern across the globe. At present, direct-acting antivirals are the treatment of choice. However, the long-term effect of this therapy has yet to be ascertained. Previously, fluoroquinolones have been reported to inhibit HCV replication by targeting NS3 protein. Therefore, it is logical to hypothesize that the natural analogs of fluoroquinolones will exhibit NS3 inhibitory activity with substantially lesser side effects.MethodIn this study, we tested the application of a recently devised integrated in-silico Cheminformatics-Molecular Docking approach to identify physicochemically similar natural analogs of fluoroquinolones from the available databases (Ambinter, Analyticon, Indofines, Specs, and TimTec). Molecular docking and ROC curve analyses were performed, using PatchDock and Graphpad software, respectively, to compare and analyze drug-protein interactions between active natural analogs, Fluoroquinolones, and HCV NS3 protein.ResultIn our analysis, we were able to shortlist 18 active natural analogs, out of 10,399, that shared physicochemical properties with the template drugs (fluoroquinolones). These analogs showed comparable binding efficacy with fluoroquinolones in targeting 32 amino acids in the HCV NS3 active site that are crucial for NS3 activity. Our approach had around 80 % sensitivity and 70 % specificity in identifying physicochemically similar analogs of fluoroquinolones.ConclusionOur current data suggest that our approach can be efficiently applied to identify putative HCV drug inhibitors that can be taken for in vitro testing. This approach can be applied to discover physicochemically similar analogs of virtually any drug, thus providing a speedy and inexpensive approach to complement drug discovery and design, which can tremendously economize on time and money spent on the screening of putative drugs.     

A High-Performance and Flexible Chemical Structure & Data Search Engine Built on CouchDB & ElasticSearch

Computer-assisted chemical structure searching plays a critical role for efficient structure screening in cheminformatics. We designed a high-performance chemical structure & data search engine called DCAIKU, built on CouchDB and ElasticSearch engines. DCAIKU converts the chemical structure similarity search problem into a general text search problem to utilize off-the-shelf full-text search engines. DCAIKU also supports flexible document structures and heterogeneous datasets with the help of schema-less document database. Our evaluations show that DCAIKU can handle both keyword search and structural search against millions of records with both high accuracy and low latency. We expect that DCAIKU will lay the foundation towards large-scale and cost-effective structural search in materials science and chemistry research.     

Design of Angiotensin‐converting Enzyme 2 (ACE2) Inhibitors by Virtual Lead Optimization and Screening

A group of presumed drug‐like molecules that possess high in silico affinity for angiotensin‐converting enzyme 2 were computationally designed. This enzyme is a promising new target in both cardiorenal disease and some coronavirus infections. A set of substrate analogous molecules were optimized by means of the LeapFrog module of the SYBYL package. Later, Molinspiration and Molsoft were used for screening out the compounds with low oral bioavailability. Similarly, OSIRIS was used for screening out the compounds having serious side effects. At the end of several stages of screening, seven candidates to anti‐viral drugs fulfiling all the evaluated criteria were obtained. They are amenable for future studies in vitro and in vivo. These designed ligands were finally evaluated by Quantitative Structure Activity Relationship studies. 21 molecules were used to carry out the qsar models. Fom these four molecules were taken as external sets yielding models with q² = 0.652 and r² = 0.962 values.     

Simulation-based cheminformatic analysis of organelle-targeted molecules: lysosomotropic monobasic amines

Cell-based molecular transport simulations are being developed to facilitate exploratory cheminformatic analysis of virtual libraries of small drug-like molecules. For this purpose, mathematical models of single cells are built from equations capturing the transport of small molecules across membranes. In turn, physicochemical properties of small molecules can be used as input to simulate intracellular drug distribution, through time. Here, with mathematical equations and biological parameters adjusted so as to mimic a leukocyte in the blood, simulations were performed to analyze steady state, relative accumulation of small molecules in lysosomes, mitochondria, and cytosol of this target cell, in the presence of a homogenous extracellular drug concentration. Similarly, with equations and parameters set to mimic an intestinal epithelial cell, simulations were also performed to analyze steady state, relative distribution and transcellular permeability in this non-target cell, in the presence of an apical-to-basolateral concentration gradient. With a test set of ninety-nine monobasic amines gathered from the scientific literature, simulation results helped analyze relationships between the chemical diversity of these molecules and their intracellular distributions.     

A systems chemical biology study of malate synthase and isocitrate lyase inhibition in Mycobacterium tuberculosis during active and NRP growth

The ability of Mycobacterium tuberculosis (Mtb) to survive in low oxygen environments enables the bacterium to persist in a latent state within host tissues. In vitro studies of Mtb growth have identified changes in isocitrate lyase (ICL) and malate synthase (MS) that enable bacterial persistence under low oxygen and other environmentally limiting conditions. Systems chemical biology (SCB) enables us to evaluate the effects of small molecule inhibitors not only on the reaction catalyzed by malate synthase and isocitrate lyase, but the effect on the complete tricarboxylic acid cycle (TCA) by taking into account complex network relationships within that system.To study the kinetic consequences of inhibition on persistent bacilli, we implement a systems-chemical biology (SCB) platform and perform a chemistry-centric analysis of key metabolic pathways believed to impact Mtb latency. We explore consequences of disrupting the function of malate synthase (MS) and isocitrate lyase (ICL) during aerobic and hypoxic non-replicating persistence (NRP) growth by using the SCB method to identify small molecules that inhibit the function of MS and ICL, and simulating the metabolic consequence of the disruption.Results indicate variations in target and non-target reaction steps, clear differences in the normal and low oxygen models, as well as dosage dependent response. Simulation results from singular and combined enzyme inhibition strategies suggest ICL may be the more effective target for chemotherapeutic treatment against Mtb growing in a microenvironment where oxygen is slowly depleted, which may favor persistence.