The advancements of information technology and related processing techniques have created a fertile base for progress in many scientific fields and industries. In the fields of drug discovery and development, machine learning techniques have been used for the development of novel drug candidates. The methods for designing drug targets and novel drug discovery now routinely combine machine learning and deep learning algorithms to enhance the efficiency, efficacy, and quality of developed outputs. The generation and incorporation of big data, through technologies such as high-throughput screening and high through-put computational analysis of databases used for both lead and target discovery, has increased the reliability of the machine learning and deep learning incorporated techniques. The use of these virtual screening and encompassing online information has also been highlighted in developing lead synthesis pathways. In this review, machine learning and deep learning algorithms utilized in drug discovery and associated techniques will be discussed. The applications that produce promising results and methods will be reviewed.

We present a new de novo transcriptome assembler, Bridger, which takes advantage of techniques employed in Cufflinks to overcome limitations of the existing de novo assemblers. When tested on dog, human, and mouse RNA-seq data, Bridger assembled more full-length reference transcripts while reporting considerably fewer candidate transcripts, hence greatly reducing false positive transcripts in comparison with the state-of-the-art assemblers. It runs substantially faster and requires much less memory space than most assemblers. More interestingly, Bridger reaches a comparable level of sensitivity and accuracy with Cufflinks. Bridger is available at https://sourceforge.net/projects/rnaseqassembly/files/?source=navbar.

Computed tomography (CT) examinations are commonly used to predict lung nodule malignancy in patients, which are shown to improve noninvasive early diagnosis of lung cancer. It remains challenging for computational approaches to achieve performance comparable to experienced radiologists. Here we present NoduleX, a systematic approach to predict lung nodule malignancy from CT data, based on deep learning convolutional neural networks (CNN). For training and validation, we analyze >1000 lung nodules in images from the LIDC/IDRI cohort. All nodules were identified and classified by four experienced thoracic radiologists who participated in the LIDC project. NoduleX achieves high accuracy for nodule malignancy classification, with an AUC of ~0.99. This is commensurate with the analysis of the dataset by experienced radiologists. Our approach, NoduleX, provides an effective framework for highly accurate nodule malignancy prediction with the model trained on a large patient population. Our results are replicable with software available at http://bioinformatics.astate.edu/NoduleX.

High-throughput RNA-seq technology has provided an unprecedented opportunity to reveal the very complex structures of transcriptomes. However, it is an important and highly challenging task to assemble vast amounts of short RNA-seq reads into transcriptomes with alternative splicing isoforms. In this study, we present a novel de novo assembler, BinPacker, by modeling the transcriptome assembly problem as tracking a set of trajectories of items with their sizes representing coverage of their corresponding isoforms by solving a series of bin-packing problems. This approach, which subtly integrates coverage information into the procedure, has two exclusive features: 1) only splicing junctions are involved in the assembling procedure; 2) massive pell-mell reads are assembled seemingly by moving a comb along junction edges on a splicing graph. Being tested on both real and simulated RNA-seq datasets, it outperforms almost all the existing de novo assemblers on all the tested datasets, and even outperforms those ab initio assemblers on the real dog dataset. In addition, it runs substantially faster and requires less memory space than most of the assemblers. BinPacker is published under GNU GENERAL PUBLIC LICENSE and the source is available from: http://sourceforge.net/projects/transcriptomeassembly/files/BinPacker_1.0.tar.gz/download. Quick installation version is available from: http://sourceforge.net/projects/transcriptomeassembly/files/BinPacker_binary.tar.gz/download. The availability of RNA-seq technology drives the development of algorithms for transcriptome assembly from very short RNA sequences. However, the problem of how to (de novo) assemble transcriptome using RNA-seq datasets has not been modeled well; e.g. sequence coverage information has even not been accurately and effectively integrated into the appropriate assembling procedure, leading to a bottleneck that all the existing (de novo) strategies have encountered. We present a novel approach to remodel the problem as tracking a set of trajectories of items with their sizes representing the coverage of their corresponding isoforms by solving a series of bin-packing problems. This approach, which subtly integrates the coverage information into the procedure, has two exclusive features: 1) only splicing junctions are involved in the assembling procedure; 2) massive pell-mell reads are assembled seemingly by moving a comb along junction edges on a splicing graph. Being tested on both real and simulated RNA-seq datasets, it outperforms almost all existing de novo assemblers on all the tested datasets, even outperforms those ab initio assemblers on the dog dataset, in terms of commonly used comparison standards.

Biclustering algorithms, which aim to provide an effective and efficient way to analyze gene expression data by finding a group of genes with trend-preserving expression patterns under certain conditions, have been widely developed since Morganet al.pioneered a work about partitioning a data matrix into submatrices with approximately constant values. However, the identification of general trend-preserving biclusters which are the most meaningful substructures hidden in gene expression data remains a highly challenging problem. We found an elementary method by which biologically meaningful trend-preserving biclusters can be readily identified from noisy and complex large data. The basic idea is to apply the longest common subsequence (LCS) framework to selected pairs of rows in an index matrix derived from an input data matrix to locate a seed for each bicluster to be identified. We tested it on synthetic and real datasets and compared its performance with currently competitive biclustering tools. We found that the new algorithm, named UniBic, outperformed all previous biclustering algorithms in terms of commonly used evaluation scenarios except for BicSPAM on narrow biclusters. The latter was somewhat better at finding narrow biclusters, the task for which it was specifically designed.

We present here the Arkansas AI-Campus solution method for the 2019 Kidney Tumor Segmentation Challenge (KiTS19). Our Arkansas AI-Campus team participated the KiTS19 Challenge for four months, from March to July of 2019. This paper provides a summary of our methods, training, testing and validation results for this grand challenge in biomedical imaging analysis. Our deep learning model is an ensemble of U-Net models developed after testing many model variations. Our model has consistent performance on the local test dataset and the final competition independent test dataset. The model achieved local test Dice scores of 0.949 for kidney and tumor segmentation, and 0.601 for tumor segmentation, and the final competition test earned Dice scores 0.9470 and 0.6099 respectively. The Arkansas AI-Campus team solution with a composite DICE score of 0.7784 has achieved a final ranking of top fifty worldwide, and top five among the United States teams in the KiTS19 Competition.