This special issue of the Journal of Computational Biology includes a selection of articles presented at the 11th International Computational Advances in Bio and Medical Sciences (ICCABS 2021), which was held in a virtual format during December 16–18, 2021. ICCABS has the goal of bringing together researchers, scientists, and students from academia, laboratories, and industry to discuss recent advances on computational techniques and applications in the areas of biology, medicine, and drug discovery.
In 2021, 13 extended abstracts were submitted in response to the call for papers, out of which 10 extended abstracts were accepted for oral presentation at the conference and publication in the ICCABS 2021 postproceedings published as volume 13,254 of Springer Verlag's Lecture Notes in Bioinformatics series.
The technical program of ICCABS 2021 also included 9 invited talks presented at the 11th Workshop on Computational Advances for Next Generation Sequencing (CANGS 2021), 16 invited talks presented at the 10th Workshop on Computational Advances in Molecular Epidemiology (CAME 2021), 12 invited talks presented at the 4th Workshop on Computational Advances for Single-Cell Omics Data Analysis (CASCODA 2021), and 7 invited talks presented at the 1st Workshop on Advances in Systems Immunology (ASI 2021).
This special issue includes a selection of 12 articles presented at ICCABS 2021 and the joint CANGS, CAME, CASCODA, and ASI workshops. The first article, by Mitchell et al., was selected from the extended abstracts presented at the main ICCABS 2021 conference. In this article, the authors introduce the excerno R package for filtering false variant calls induced by DNA degradation in formalin-fixed paraffin-embedded (FFPE) samples. The package integrates the known signature of FFPE artifacts within a Bayesian framework to annotate mutations called from sequencing data generated from FFPE samples, which are commonly used in cancer genomics studies. The specificity and sensitivity of excerno are assessed using data sets simulating varying degree of FFPE-induced DNA degradation and mutations generated by different mutational processes captured from the Catalogue of Somatic Mutations in Cancer (COSMIC) databases.
The next three articles were selected from the invited talks presented at CANGS 2021. The article by Li et al. describes a novel bimodal susceptible, exposed, infectious, and removed model that uses different epidemiological parameters for the local community and workplace. The authors evaluated the model by extensive simulations and successfully used it to guide COVID-19 testing strategies at the Jackson Laboratory. The article by Tithi et al. describes a new bioinformatic pipeline, called FVE-novel, for reconstructing viral genome sequences from metagenomic sequencing data. The pipeline relies on the previously developed FastViromeExplorer tool to select reads that align to reference viral genomes then performs iterative de novo assembly of the selected reads. Running the pipeline on an ocean metagenomic data set produced four complete viral genomes, two of which are novel.
In the third of these three articles, Tan et al. present a novel method, called major minor variation clustering (MMVC), for identifying polio outbreaks. MMVC uses a network model to simultaneously incorporate sequence similarity in major and minor viral variants along with epidemiological information such as onset dates. Analysis of a public data set shows that MMVC produces improved clustering results compared with a phylogenetic tree approach.
The next six articles were selected from the invited talks presented at CAME 2021. The article by Campo et al. introduces a new information-theoretic distance called Mutual Information and Entropy H (MIH) for nucleotide or amino acid sequences. MIH is shown to outperform other common distances in identifying pairs of cross-immunoreactive hepatitis C virus sequences and peptides that bind to the major histocompatibility complex. MIH is also shown to better differentiate between RNA sequences with different secondary structures, suggesting its utility as a proxy for phenotypic similarity in a broad range of scenarios.
The article by Ali et al. presents scalable alignment-free machine learning methods for classifying spike protein sequences of SARS-CoV-2 based on their geographical location. The authors show that their approach significantly outperforms existing analysis pipelines and analyze the information gain associated with different amino acid positions in the SARS-CoV-2 spike protein.
The article by Senchyna and Singh explores temporal network representations of SARS-CoV-2 sequencing data as an alternative to standard phylogenetic trees. The authors characterize global and local properties of these networks, and show their utility by analyzing SARS-CoV-2 data sets from Israel and France.
The article by Chourasia et al. introduces an embedding technique called Reads2Vec that enables the analysis of viral sequencing data sets without assembly or alignment. The authors show that Reads2Vec outperforms previously proposed alignment-free embedding methods in classification and clustering accuracy on simulated and real SARS-CoV-2 data sets, and has much faster running time than the alignment-based pangolin pipeline.
In their article, Bunimovich and Ram explore network models of cross-immunoreactivity between viral variants with respect to mechanisms of antigenic cooperation that could explain the phenomenon of local immunodeficiency and chronic infections. Among others, they show that connecting simple cross-immunoreactivity networks that have a stable state of local immunodeficiency may result in virus variants changing roles and the existence of subnetworks that evolve independently of each other.
Finally, Sahoo et al. use a support vector machine classifier with recursive feature elimination to generate a set of 24 genes that differentiate with high accuracy between triple-negative breast cancer (TNBC) samples of African American women and those of European American women. The selected genes include KLK10, a karllikrein-related peptidase whose expression has been previously associated with progression of ovarian and other cancers. The authors find that for TNBC, high expression of KLK10 is associated with worse survival, potentially explaining some of the observed racial disparities in survival rates of TNBC patients.
The last two articles were selected from the invited talks presented at CASCODA 2021 and ASI 2021. Moravec et al. investigate in their article the feasibility of constructing tumor phylogenies from single-cell RNA-Seq (scRNA-Seq) data. The authors show that, despite the low and uneven coverage of scRNA-Seq data, it is possible to identify sufficient single-nucleotide variants (SNVs) for reconstructing robust tumor phylogenies. SNV-based phylogenies are found to be similar to those reconstructed from normalized gene expression levels, suggesting that the reconstructed phylogenies reflect the true clonal evolution of the tumors.
The last article in this special issue, by Al Seesi et al., presents GeNeo, a comprehensive set of Galaxy tools for identification of tumor-specific neoepitopes from matched tumor/normal exome sequencing data coupled with tumor RNA-Seq data. Among others, GeNeo includes tools for somatic variant calling from multi-technology sequencing data generated using the Illumina and Ion Torrent platforms, variant validation by targeted resequencing using the AccessArray platform, and neoepitope prediction and filtering. The authors also detail the application of these tools in a phase I clinical trial.
We thank the Editor-in-Chief, Mona Singh, for providing us the opportunity to highlight some of the exciting research presented at ICCABS 2021 in the Journal of Computational Biology. Finally, we thank all ICCABS authors—the conference could not continue to thrive without their high-quality contributions.