Irtesam Mahmud Khan

Single-cell RNA sequencing (scRNA-seq) has transformed transcriptome profiling at cellular resolution, yet accurate reconstruction of full-length transcripts for individual cells remains a central challenge. Emerging scRNA-seq protocols can produce reads that span entire transcripts, enabling isoform-level expression analysis. For example, Smart-seq protocols combine unique molecular identifier (UMI)-linked reads that index and stitch together multiple reads from the same molecule, with internal reads filling coverage gaps. We demonstrate that these read types exhibit markedly different biological and statistical properties in strandness, 5'/3' coverage bias, and genomic locality. Existing assemblers fail to leverage these distinctions, yielding suboptimal assembly. We developed Amaranth, a novel single-cell assembler that discriminatively models UMI and internal reads. Amaranth implements heuristics specifically designed to address the distinct biases of UMI-linked and internal reads, enabling accurate strandness assignment for internal reads, reliable splicing graph refinement, and precise transcript start site determination. We also developed Amaranth-meta, which integrates information across cells to enhance individual cell assemblies. Benchmarked on Smart-seq3 datasets from human HEK293T and mouse fibroblast cells, Amaranth outperformed other state-of-the-art assemblers in assembling individual cells and in meta-assembly. Amaranth advances isoform-level analysis in single-cell transcriptomics, facilitating detailed studies at cellular resolution. Amaranth is implemented in C++ and is freely available at https://github.com/Shao-Group/amaranth under the BSD-3-Clause license. Scripts, documentation, and data for reproducing experiments in this manuscript are available at https://github.com/Shao-Group/amaranth-test.

Transcriptome annotations remain incomplete despite enormous efforts. Annotations are largely driven by experimental data, whereas little is understood from an evolutionary perspective. Here we present TENNIS, a model for isoform representation and inference. TENNIS models isoforms in a transcript group as nodes of a connected graph, in which the edges represent basic alternative splicing events, and predicts missing isoforms using a novel algorithm. Our analysis indicates that approximately 80% of the analyzed isoform groups satisfy our model, whereas the identified missing transcripts show high accuracy. TENNIS achieves these results without using additional sequencing data, offering insights into alternative splicing and a powerful tool for constructing annotations.

Motivation: Single-cell RNA sequencing has transformed transcriptome profiling at cellular resolution, yet accurate reconstruction of full-length transcripts for individual cells remains a central challenge. Emerging scRNA-seq protocols can produce reads that span entire transcripts, enabling isoform-level expression analysis. For example, Smart-seq protocols combine UMI-linked reads that index and stitch together multiple reads from the same molecule, with internal reads filling coverage gaps. We demonstrate that these read types exhibit markedly different biological and statistical properties in strandness, 5’/3’ coverage bias, and genomic locality. Existing assemblers fail to leverage these distinctions, yielding suboptimal assembly. Results: We developed Amaranth, a novel single-cell assembler that discriminatively models UMI and internal reads. Amaranth implements heuristics specifically designed to address the distinct biases of UMI-linked and internal reads, enabling accurate strandness assignment for internal reads, reliable splicing graph refinement, and precise transcript start site determination. We also developed Amaranth-meta, which integrates information across cells to enhance individual cell assemblies. Benchmarked on Smart-seq3 datasets from human HEK293T and mouse fibroblast cells, Amaranth outperformed other state-of-the-art assemblers in assembling individual cells and in meta-assembly. Amaranth advances isoform-level analysis in single-cell transcriptomics, facilitating detailed studies at cellular resolution. Availability and Implementation: Amaranth is implemented in C++ and is freely available at https://github.com/Shao-Group/amaranth under the BSD-3-Clause license. Scripts, documentation, and data for reproducing experiments in this manuscript are available at https://github.com/Shao-Group/amaranth-test .

Rolling circle amplification (RCA) coupled with long-read sequencing has emerged as a powerful strategy for detecting full-length circular RNAs (circRNAs). Such protocols produce long reads that are normally composed of several tandemly repeated copies of the original circRNA. The circRNA sequence can be reconstructed through detecting the repeating unit of the long reads, which are aligned to the genome to validate and to identify back-splice junctions (BSJs). We revealed a previously unrecognized phenomenon: a substantial fraction of long reads contain complex repeat patterns in which the repeating unit consists of a sequence combined with its partial reverse complement. In these cases, only the original sequence corresponds to the true circRNA, while the concatenated pattern may produce false circRNAs and misidentify correct circRNAs. We present a new approach CircPlex that extracts the authentic circRNA sequence from these complex repeat units, overcoming the limitations of standard repeat-based consensus prediction. Comparison with isoCirc annotations and circRNA database demonstrates that a significant number of BSJs and full-length sequences, previously ignored, can be recovered. Our results suggest that leveraging partially repetitive reads from RCA-based sequencing can substantially increase circRNA detection sensitivity and uncover novel isoforms, providing a more comprehensive view of the circular transcriptome.

Transcript assembly remains a challenging task despite the development of numerous methods. A major contributor to low assembly accuracy is the difficulty in accurately determining transcript start sites (TSSs) and end sites (TESs), due to the weak and noisy signals typically found in RNA-seq data. We present Telos, a two-stage machine learning framework for precise detection of TSSs and TESs and for transcript ranking. The method takes as input any assembly, typically generated by an existing assembler. In the first stage, Telos scores the TSSs and TESs in the input assembly using a machine learning model trained on a rich set of engineered features. These site-level scores will be passed to the second stage for transcript-level evaluation. In its second stage, Telos scores the entire transcripts by training another model that integrates features of their TSS and TES (including the inferred probabilities from the first stage), along with transcript abundance estimated by the assembler and statistics about exon lengths. We extensively evaluated Telos on ONT (cDNA and direct RNA), PacBio, and Illumina short-read RNA-seq datasets. In all cases, it consistently outperformed baseline methods. Telos is agile, but achieves substantial improvements, demonstrating the value of explicitly modeling TSS and TES, a gap in current transcript assembly tools. Telos can be paired with any assembler to accurately score the assembled transcripts. It is modular, easily extensible to emerging sequencing technologies, and hence we anticipate its broad adoption in transcriptomic studies.

Alternative splicing (AS) is a ubiquitous mechanism in eukaryotes. It is estimated that 90% of human genes are alternatively spliced. Despite enormous efforts, transcriptome annotations remain, nevertheless, incomplete. Conventional means of annotation were largely driven by experimental data such as RNA-seq and protein sequences, while little insight was shed on understanding transcriptomes and alternative splicings from the perspective of evolution. This study addresses this critical gap by presenting TENNIS (Transcript EvolutioN for New Isoform Splicing), an evolution-based model to predict unannotated isoforms and refine existing annotations without requiring additional data. The model of TENNIS is based on two minimal premises–AS isoforms evolve sequentially from existing isoforms, and each evolutionary step involves a single AS event. We formulate the identification of missing transcripts as an optimization problem and parsimoniously find the minimal number of novel transcripts. Our analysis showed approximately 80% of multi-transcript groups from six transcriptome annotations satisfy our evolutionary model. At a high confidence level, 40% of isoforms predicted by TENNIS were validated by deep long-read RNA-seq. In a simulated incomplete annotation scenario, TENNIS dramatically outperforms two randomized baseline approaches by a 2.25-3 fold-change in precision or a 3.5-3.9 fold-change in recall, after controlling the same level of recall or precision of the baseline methods. These results demonstrate that TENNIS effectively identifies missing transcripts by complying with minimal propositions, offering a powerful approach for transcriptome augmentations through the lens of alternative splicing evolutions. TENNIS is freely available at https://github.com/Shao-Group/tennis .

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus responsible for coronavirus disease 2019 (COVID-19), has had an unprecedented effect, especially among under-resourced minority communities. Surveillance of those at high risk is critical for preventing and controlling the pandemic. We must better understand the relationships between COVID-19-related cases or deaths and characteristics in our most vulnerable population that put them at risk to target COVID-19 prevention and management efforts. Population characteristics strongly related to United States (US) county-level data on COVID-19 cases and deaths during all stages of the pandemic were identified from the onset of the epidemic and included county-level socio-demographic and comorbidities data, as well as daily meteorological modeled observation data from the North American Regional Reanalysis (NARR), and the NARR high spatial resolution model to assess the environment. Advanced machine learning (ML) approaches were used to identify outbreaks (geographic clusters of COVID-19) and included spatiotemporal risk factors and COVID-19 vaccination efforts, especially among vulnerable and underserved communities. COVID-19 outcomes were found to be negatively associated with the number of people vaccinated and positively associated with age, the prevalence of cardiovascular disease, diabetes, and the minority population. There was also a strong positive correlation between unauthorized immigrants and the prevalence of COVID-19 cases and deaths. Meteorological variables were also investigated, but correlations with COVID-19 were relatively weak. Our findings suggest that COVID-19 has had a disproportionate impact across the US population among vulnerable and minority communities. Findings also emphasize the importance of vaccinations and tailored public health initiatives (e.g., mask mandates, vaccination) to reduce the spread of COVID-19 and the number of COVID-19 related deaths across all populations.

The COVID-19 pandemic has spread globally. Only three cases in Bangladesh were reported on March 8, 2020. Here, we aim to predict the epidemic progression for 1 year under different scenarios in Bangladesh. We extracted the number of daily confirmed cases from March 8 to July 20, 2020. We considered the suspected-infected-removed (SIR) model and performed a maximum likelihood-based grid search to determine the removal rate (ɣ). The transmission was modeled as a stochastic random walk process, and sequential Monte Carlo simulation was run 100 times with bootstrap fits to infer the transmission rate (β) and R t . According to the simulation, the (real) peak daily incidence of 3,600 would be followed by a steady decline, reaching below 1,000 in late January 2021. Thus, the model predicted that there would still be more than 300 cases/day even after a year. However, with proper interventions, a much steeper decline would be achieved following the peak. If we apply a combined (0.8β, 1.2ɣ) intervention, there would be less than 100 cases by mid-October, only around five odd cases at the beginning of the year 2021, and zero cases in early March 2021. The predicted total number of deaths (in status quo) after 1 year would be 8,533 which would reduce to 3,577 if combined (0.8β, 1.2ɣ) intervention is applied. We have also predicted the ideal number of tests that Bangladesh should perform and based on that redid the whole simulation. The outcome, though worse, would be manageable with interventions according to the simulation.

Background: The COVID-19 epidemic had spread rapidly through China and subsequently has proliferated globally leading to a pandemic situation around the globe. Human-to-human transmissions, as well as asymptomatic transmissions of the infection, have been confirmed. As of April 3^rd, public health crisis in China due to COVID-19 is potentially under control. Methods: We compiled a daily dataset of case counts, mortality, recovery, temperature, population density, and demographic information for each prefecture during the period of January 11 to April 07, 2020 (excluding Wuhan from our analysis due to missing data). Understanding the characteristics of spatiotemporal clustering of the COVID-19 epidemic and R₀ is critical in effectively preventing and controlling the ongoing global pandemic. The prefectures were grouped based on several relevant features using unsupervised machine learning techniques. We performed a computational analysis utilizing the reported cases in China to estimate the revised R₀ among different regions for prevention planning in an ongoing global pandemic. Results: Finally, our results indicate that the impact of temperature and demographic (different age group percentage compared to the total population) factors on virus transmission may be characterized using a stochastic transmission model. Conclusions: Such predictions will help prioritize segments of a given community/ region for action and provide a visual aid in designing prevention strategies for a specific geographic region. Furthermore, revised estimation and our methodology will aid in improving the human health consequences of COVID-19 elsewhere.