Biology subjects

Hajirasouliha, I.

Publications and source records attributed to Hajirasouliha, I..

At least 19 recordsLinked to original sources

bioRxiv2025

Spatial isoform sequencing at sub-micrometer single-cell resolution reveals novel patterns of spatial isoform variability in brain cell types

Spatial long-read technologies are becoming increasingly common but lack nanometer and therefore often single-cell resolution. This leaves the question unanswered of whether spatially variable isoforms represent spatial variability within one cell type or differences in cell-type composition between different regions. Here, we developed Spl-ISO-Seq2 (220nm spot size and 500nm resolution), and the accompanying software packages Spl-IsoQuant-2 and Spl-IsoFind, enabling long-read sequencing using up to 500 million barcodes compared to 80,000 previously. Applying this to the adult mouse brain, we compared spatial variability by examining (a) differential isoform abundance between known brain regions and (b) spatial isoform patterns that do not align with predefined regions. While the former revealed more spatial isoform differences, both approaches identified overlapping hits, e.g., Rps24 in oligodendrocytes. For Snap25, previously known to exhibit spatial isoform variation, we now show that this variability occurs in excitatory neurons. The second approach also uncovered patterns not captured by predefined-region comparisons, e.g., Tnnc1 in excitatory neurons. Furthermore, we show that a surprising number of spatial isoform signals is not driven by cell-type composition alone. Finally, we applied our software to newly generated Visium HD 3 long-read data to demonstrate its applicability and strong reproducibility across protocols and biological replicates. Taken together, our experimental and analytical methods enrich spatial transcriptomics with a so-far elusive isoform view of spatial variation for individual cell types.

neuroscience↗

bioRxiv2025

A spatial long-read approach at near-single-cell resolution reveals developmental regulation of splicing and polyadenylation sites in distinct cortical layers and cell types.

Genome-wide single-cell and spatial long-read approaches have gained traction, but mostly lack single-cell resolution - and yield limited read lengths. Here, we introduce spatial ISOform sequencing (Spl-ISO-Seq), which reveals exons and polyadenylation sites from long reads with near-single-cell resolution. Spl-ISO-Seq selects long cDNAs and doubles to triples read lengths compared to standard preparations. Adding a highly specific software tool (Spl-ISOquant) and comparing human post-mortem pre-puberty samples of the visual cortex (8-11 years) to post-puberty samples (16-19 years), we find that cortical layers harbor stronger splicing and poly(A)-site regulation than the adjacent white matter, with enrichment of multiple protein-domain types. For oligodendrocytes however, developmental splicing changes are stronger in white matter. Among cortical layers, layer 4 has the most developmental changes in alternative-exon inclusion in excitatory neurons and in poly(A) sites. We also find many repeat elements, especially ERV1 long terminal repeats downstream of developmentally-regulated layer 4 exons. Overall, alternative splicing changes are linked to synapses - specifically at the post-synapse. Age-linked splicing changes in layers 1-3 and 4 are associated with autism spectrum disorder but not with schizophrenia, amyotrophic lateral sclerosis and Alzheimers disease. These results root developmental splicing changes during puberty and the resulting protein changes in specific layers and cell types. More generally, our new technologies enable new observations for any complex tissue.

genomics↗

bioRxiv2024

Blackbird: structural variant detection using synthetic and low-coverage long-reads

MotivationRecent benchmarks of structural variant (SV) detection tools revealed that the majority of human genome structural variations (SVs), especially the medium-range (50-10,000 bp) SVs cannot be re-solved with short-read sequencing, but long-read SV callers achieve great results on the same datasets. While improvements have been made, high-coverage long-read sequencing is associated with higher costs and input DNA requirements. To decrease the cost one can lower the sequence coverage, but the current long-read SV callers perform poorly with coverage below 10x. Synthetic long-read (SLR) technologies hold great potential for structural variant (SV) detection, although utilizing their long-range information for events smaller than 50 kbp has been challenging. ResultsIn this work, we propose a hybrid novel integrated alignment- and local-assembly-based algorithm, Blackbird, that uses SLR together with low-coverage long reads to improve SV detection and assembly. Without the need for a computationally expensive whole genome assembly, Blackbird uses a sliding window approach and barcode information encoded in SLR to accurately assemble small segments and use long reads for an improved gap closing and contig assembly. We evaluated Blackbird on simulated and real human genome datasets. Using the HG002 GIAB benchmark set, we demonstrated that in hybrid mode, Blackbird demonstrated results comparable to state-of-the-art long-read tools, while using less long-read coverage. Blackbird requires only 5x coverage to achieve F1 scores (0.835 and 0.808 for deletions and insertions) similar to PBSV (0.856 and 0.812) and Sniffles2 (0.839 and 0.804) using 10x Pacbio Hi-Fi long-read coverage.

bioinformatics↗

bioRxiv2023

Automatic Ploidy Prediction and Quality Assessment of Human Blastocyst Using Time-Lapse Imaging

Assessing fertilized human embryos is crucial for in vitro-fertilization (IVF), a task being revolutionized by artificial intelligence and deep learning. Existing models used for embryo quality assessment and chromosomal abnormality (ploidy) detection could be significantly improved by effectively utilizing time-lapse imaging to identify critical developmental time points for maximizing prediction accuracy. Addressing this, we developed and compared various embryo ploidy status prediction models across distinct embryo development stages. We present BELA (Blastocyst Evaluation Learning Algorithm), a state-of-the-art ploidy prediction model surpassing previous image- and video-based models, without necessitating subjective input from embryologists. BELA uses multitask learning to predict quality scores that are used downstream to predict ploidy status. By achieving an AUC of 0.76 for discriminating between euploidy and aneuploidy embryos on the Weill Cornell dataset, BELA matches the performance of models trained on embryologists manual scores. While not a replacement for preimplantation genetic testing for aneuploidy (PGT-A), BELA exemplifies how such models can streamline the embryo evaluation process, reducing time and effort required by embryologists.

developmental biology↗

bioRxiv2023

cloudrnaSPAdes: Isoform assembly using bulk barcoded RNA sequencing data

MotivationRecent advancements in long-read RNA sequencing have enabled the examination of full-length isoforms, previously uncaptured by short-read sequencing methods. An alternative powerful method for studying isoforms is through the use of barcoded short-read RNA reads, for which a barcode indicates whether two short-reads arise from the same molecule or not. Such techniques included the 10x Genomics linked-read based SParse Isoform Sequencing (SPIso-seq), as well as Loop-Seq, or Tell-Seq. Some applications, such as novel-isoform discovery, require very high coverage. Obtaining high coverage using long reads can be difficult, making barcoded RNA-seq data a valuable alternative for this task. However, most annotation pipelines are not able to work with a set of short reads instead of a single transcript, also not able to work with coverage gaps within a molecule if any. In order to overcome this challenge, we present an RNA-seq assembler allowing the determination of the expressed isoform per barcode. ResultsIn this paper, we present cloudrnaSPAdes, a tool for assembling full-length isoforms from barcoded RNA-seq linked-read data in a reference-free fashion. Evaluating it on simulated and real human data, we found that cloudrnaSPAdes accurately assembles isoforms, even for genes with high isoform diversity. AvailabilitycloudrnaSPAdes is a feature release of a SPAdes assembler and available at https://cab.spbu.ru/software/cloudrnaspades/. Contactdmm2017@med.cornell.edu

bioinformatics↗

bioRxiv2023

A modular metagenomics analysis system for integrated multi-step data exploration

MotivationComputational analysis of large-scale metagenomics sequencing datasets have proven to be both incredibly valuable for extracting isolate-level taxonomic, and functional insights from complex microbial communities. However, due to an ever-expanding ecosystem of metagenomics-specific methods and file-formats, designing seamless and scalable end-to-end workflows, and exploring the massive amounts of output data have become studies unto themselves. One-click bioinformatics pipelines have helped to organize these tools into targeted workflows, but they suffer from general compatibility and maintainability issues, and preclude replication. MethodsTo address the gap in easily extensible yet robustly distributable metagenomics workflows, we have developed a module-based metagenomics analysis system "Core Analysis Modular Pipeline" (CAMP), written in Snakemake, a popular workflow management system, along with a standardized module and working directory architecture. Each module can be run independently or conjointly with a series of others to produce the target data format (e.g. short-read preprocessing alone, or short-read preprocessing followed by de novo assembly), and outputs aggregated summary statistics reports and semi-guided Jupyter notebook-based visualizations. ResultsWe have applied CAMP to a set of ten metagenomics samples to demonstrate how a modular analysis system with built-in data visualization at intermediate steps facilitates rich and seamless inter-communication between output data from different analytic purposes. AvailabilityThe CAMP ecosystem (module template and analysis modules) can be found https://github.com/Meta-CAMP.

bioinformatics↗

bioRxiv2022

A deep learning framework for structural variant discovery and genotyping

Structural variants (SV) are a major driver of genetic diversity and disease in the human genome and their discovery is imperative to advances in precision medicine and our understanding of human genetics. Existing SV callers rely on hand-engineered features and heuristics to model SVs, which cannot easily scale to the vast diversity of SV types nor fully harness all the information available in sequencing datasets. Since deep neural networks can learn complex abstractions directly from the data, they offer a promising approach for general SV discovery. Here we propose an extensible deep learning framework, Cue, to call and genotype SVs. At a high level, Cue converts sequence alignments to multi-channel images that capture multiple SV-informative signals and uses a stacked hourglass convolutional neural network to predict the type, genotype, and genomic locus of the SVs captured in each image. We show that Cue outperforms the state of the art in the detection of five classes of SVs (including two types of complex SVs and subclonal SVs) on synthetic and real short-read data and that it can be easily extended to other sequencing platforms, such as long and linked read sequencing technologies, while achieving competitive performance. By design, Cue can also be automatically extended to support new SV classes: this versatility is crucial as novel SV types are discovered in ongoing population-scale sequencing initiatives.

bioinformatics↗

bioRxiv2021

Weakly-Supervised Tumor Purity Prediction FromFrozen H&E Stained Slides

Estimating tumor purity is especially important in the age of precision medicine. Purity estimates have been shown to be critical for correction of tumor sequencing results, and higher purity samples allow for more accurate interpretations from next-generation sequencing results. In addition, tumor purity has been shown to be correlated with survival outcomes for several diseases. Molecular-based purity estimates using computational approaches require sequencing of tumors, which is both time-consuming and expensive. Here we propose an approach, weakly-supervised purity (wsPurity), which can accurately quantify tumor purity within a slide, using multiple and different types of cancer. This approach allows for a flexible analysis of tumors from whole slide imaging (WSI) of histology hematoxylin and eosin (H&E) slides. Our model predicts tumor type with high accuracy (greater than 80% on an independent test cohort), and tumor purity at a higher accuracy compared to a comparable fully-supervised approach (0.1335 MAE on an independent test cohort). In addition to tumor purity prediction, our approach can identify high resolution tumor regions within a slide, to enrich tumor cell selection for downstream analyses. This model could also be used in a clinical setting, to stratify tumors into high and low tumor purity, using different thresholds, in a cancer-dependent manner, depending on what purity levels correlate with worse disease outcomes. In addition, this approach could be used in clinical practice to select the best tissue block for sequencing. Overall, this approach can be used in several different ways to analyze WSIs of tumor H&E sections.

bioinformatics↗

bioRxiv2021

Ariadne: Barcoded Linked-Read Deconvolution Using de Bruijn Graphs

Synthetic Long Read (SLR) sequencing techniques such as USTs TELL-Seq, and Loop Genomics LoopSeq combine 3' barcoding with standard short-read sequencing to expand the range of linkage resolution from hundreds to tens of thousands of base-pairs. However, the lack of a 1:1 correspondence between a long fragment and a 3' unique molecular identifier (UMI) confounds the assignment of linkage between short-reads. We introduce Ariadne, a novel assembly graph-based SLR deconvolution algorithm, that can be used to extract single-species read-clouds from SLR datasets to improve the taxonomic classification and de novo assembly of complex populations, such as metagenomes.

genomics↗

bioRxiv2021

Systemic Tissue and Cellular Disruption from SARS-CoV-2 Infection revealed in COVID-19 Autopsies and Spatial Omics Tissue Maps

The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) virus has infected over 115 million people and caused over 2.5 million deaths worldwide. Yet, the molecular mechanisms underlying the clinical manifestations of COVID-19, as well as what distinguishes them from common seasonal influenza virus and other lung injury states such as Acute Respiratory Distress Syndrome (ARDS), remains poorly understood. To address these challenges, we combined transcriptional profiling of 646 clinical nasopharyngeal swabs and 39 patient autopsy tissues, matched with spatial protein and expression profiling (GeoMx) across 357 tissue sections. These results define both body-wide and tissue-specific (heart, liver, lung, kidney, and lymph nodes) damage wrought by the SARS-CoV-2 infection, evident as a function of varying viral load (high vs. low) during the course of infection and specific, transcriptional dysregulation in splicing isoforms, T cell receptor expression, and cellular expression states. In particular, cardiac and lung tissues revealed the largest degree of splicing isoform switching and cell expression state loss. Overall, these findings reveal a systemic disruption of cellular and transcriptional pathways from COVID-19 across all tissues, which can inform subsequent studies to combat the mortality of COVID-19, as well to better understand the molecular dynamics of lethal SARS-CoV-2 infection and other viruses.

microbiology↗

bioRxiv2019

Novel Algorithms for the Taxonomic Classification of Metagenomic Linked Reads

We present KrakenLinked, a metagenomic read classifier for Linked-Reads. We have formulated two algorithms for read classification of metagenomic samples using linked reads: tree pruning and taxa promotion. Tree pruning improves specificity while taxa promotion improves sensitivity. Used together the algorithms improve taxonomic classification of linked-reads compared to short reads, particularly reducing false identification of taxa. We have implemented these algorithms as functions in KrakenUniq which we make available as KrakenLinked.

bioinformatics↗

bioRxiv2018

Robust Automated Assessment of Human Blastocyst Quality using Deep Learning

Morphology assessment has become the standard method for evaluation of embryo quality and selecting human blastocysts for transfer in in vitro fertilization (IVF). This process is highly subjective for some embryos and thus prone to human bias. As a result, morphological assessment results may vary extensively between embryologists and in some cases may fail to accurately predict embryo implantation and live birth potential. Here we postulated that an artificial intelligence (AI) approach trained on thousands of embryos can reliably predict embryo quality without human intervention.\n\nTo test this hypothesis, we implemented an AI approach based on deep neural networks (DNNs). Our approach called STORK accurately predicts the morphological quality of blastocysts based on raw digital images of embryos with 98% accuracy. These results indicate that a DNN can automatically and accurately grade embryos based on raw images. Using clinical data for 2,182 embryos, we then created a decision tree that integrates clinical parameters such as embryo quality and patient age to identify scenarios associated with increased or decreased pregnancy chance. This IVF data-driven analysis shows that the chance of pregnancy varies from 13.8% to 66.3%.\n\nIn conclusion, our AI-driven approach provides a novel way to assess embryo quality and uncovers new, potentially personalized strategies to select embryos with an improved likelihood of pregnancy outcome.

bioinformatics↗

bioRxiv2018

Characterization of segmental duplications and large inversions using Linked-Reads

Many algorithms aimed at characterizing genomic structural variation (SV) have been developed since the inception of high-throughput sequencing. However, the full spectrum of SVs in the human genome is not yet assessed. Most of the existing methods focus on discovery and genotyping of deletions, insertions, and mobile elements. Detection of balanced SVs with no gain or loss of genomic segments (e.g., inversions) is particularly a challenging task. Long read sequencing has been leveraged to find short inversions but there is still a need to develop methods to detect large genomic inversions. Furthermore, currently there are no algorithms to predict the insertion locus of large interspersed segmental duplications.\n\nHere we propose novel algorithms to characterize large (>40Kbp) interspersed segmental duplications and (>80Kbp) inversions using Linked-Read sequencing data. Linked-Read sequencing provides long range information, where Illumina reads are tagged with barcodes that can be used to assign short reads to pools of larger (30-50 Kbp) molecules. Our methods rely on split molecule sequence signature that we have previously described [11]. Similar to the split read, split molecules refer to large segments of DNA that span an SV breakpoint. Therefore, when mapped to the reference genome, the mapping of these segments would be discontinuous. We redesign our earlier algorithm, VALOR, to specifically leverage Linked-Read sequencing data to discover large inversions and characterize interspersed segmental duplications. We implement our new algorithms in a new software package, called VALOR2.\n\nAvailabilityVALOR2 is available at https://github.com/BilkentCompGen/valor.

bioinformatics↗

bioRxiv2018

PhISCS - A Combinatorial Approach for Sub-perfect Tumor Phylogeny Reconstruction via Integrative use of Single Cell and Bulk Sequencing Data

Recent technological advances in single cell sequencing (SCS) provide high resolution data for studying intra-tumor heterogeneity and tumor evolution. Available computational methods for tumor phylogeny inference via SCS typically aim to identify the most likely perfect phylogeny tree satisfying infinite sites assumption (ISA). However limitations of SCS technologies such as frequent allele dropout or highly variable sequence coverage, commonly result in mutational call errors and prohibit a perfect phylogeny. In addition, ISA violations are commonly observed in tumor phylogenies due to the loss of heterozygosity, deletions and convergent evolution. In order to address such limitations, we, for the first time, introduce a new combinatorial formulation that integrates single cell sequencing data with matching bulk sequencing data, with the objective of minimizing a linear combination of (i) potential false negatives (due to e.g. allele dropout or variance in sequence coverage) and (ii) potential false positives (due to e.g. read errors) among mutation calls, as well as (iii) the number of mutations that violate ISA - to define the optimal sub-perfect phylogeny. Our formulation ensures that several lineage constraints imposed by the use of variant allele frequencies (VAFs, derived from bulk sequence data) are satisfied. We express our formulation both in the form of an integer linear program (ILP) and - for the first time in the context of tumor phylogeny reconstruction - a boolean constraint satisfaction problem (CSP) and solve them by leveraging state-of-the-art ILP/CSP solvers. The resulting method, which we name PhISCS, is the first to integrate SCS and bulk sequencing data under the finite sites model. Using several simulated and real SCS data sets, we demonstrate that PhISCS is not only more general but also more accurate than the alternative tumor phylogeny inference tools. PhISCS is very fast especially when its CSP based variant is used returns the optimal solution, except in rare instances for which it provides an optimality gap. PhISCS is available at https://github.com/haghshenas/PhISCS.

bioinformatics↗

bioRxiv2018

gpps: An ILP-based approach for inferringcancer progression with mutation losses fromsingle cell data

cancer biology↗

bioRxiv2018

Inferring Cancer Progression from Single Cell Sequencing while allowing loss of mutations

MotivationIn recent years, the well-known Infinite Sites Assumption (ISA) has been a fundamental feature of computational methods devised for reconstructing tumor phylogenies and inferring cancer progressions seen as an accumulation of mutations. However, recent studies (Kuipers et al., 2017) leveraging Single-cell Sequencing (SCS) techniques have shown evidence of the widespread recurrence and, especially, loss of mutations in several tumor samples. Still, established methods that can infer phylogenies with mutation losses are however lacking.\n\nResultsWe present the SASC (Simulated Annealing Single-Cell inference) tool which is a new and robust approach based on simulated annealing for the inference of cancer progression from SCS data. More precisely, we introduce a simple extension of the model of evolution where mutations are only accumulated, by allowing also a limited amount of back mutations in the evolutionary history of the tumor: the Dollo-k model. We demonstrate that SASC achieves high levels of accuracy when tested on both simulated and real data sets and in comparison with some other available methods.\n\nAvailabilityThe Simulated Annealing Single-cell inference (SASC) tool is open source and available at https://github.com/sciccolella/sasc.\n\nContacts.ciccolella@campus.unimib.it

bioinformatics↗

bioRxiv2018

Breast Cancer Histopathological Image Classification: A Deep Learning Approach

Breast cancer remains the most common type of cancer and the leading cause of cancer-induced mortality among women with 2.4 million new cases diagnosed and 523,000 deaths per year. Historically, a diagnosis has been initially performed using clinical screening followed by histopathological analysis. Automated classification of cancers using histopathological images is a chciteallenging task of accurate detection of tumor sub-types. This process could be facilitated by machine learning approaches, which may be more reliable and economical compared to conventional methods.\n\nTo prove this principle, we applied fine-tuned pre-trained deep neural networks. To test the approach we first classify different cancer types using 6, 402 tissue micro-arrays (TMAs) training samples. Our framework accurately detected on average 99.8% of the four cancer types including breast, bladder, lung and lymphoma using the ResNet V1 50 pre-trained model. Then, for classification of breast cancer sub-types this approach was applied to 7,909 images from the BreakHis database. In the next step, ResNet V1 152 classified benign and malignant breast cancers with an accuracy of 98.7%. In addition, ResNet V1 50 and ResNet V1 152 categorized either benign- (adenosis, fibroadenoma, phyllodes tumor, and tubular adenoma) or malignant- (ductal carcinoma, lobular carcinoma, mucinous carcinoma, and papillary carcinoma) sub-types with 94.8% and 96.4% accuracy, respectively. The confusion matrices revealed high sensitivity values of 1, 0.995 and 0.993 for cancer types, as well as malignant- and benign sub-types respectively. The areas under the curve (AUC) scores were 0.996,0.973 and 0.996 for cancer types, malignant and benign sub-types, respectively. Overall, our results show negligible false negative (on average 3.7 samples) and false positive (on average 2 samples) results among different models. Availability: Source codes, guidelines and data sets are temporarily available on google drive upon request before moving to a permanent GitHub repository.

bioinformatics↗

bioRxiv2017

Minerva: An Alignment and Reference Free Approach to Deconvolve Linked-Reads for Metagenomics

Emerging Linked-Read technologies (aka Read-Cloud or barcoded short-reads) have revived interest in standard short-read technology as a viable way to understand large-scale structure in genomes and metagenomes. Linked-Read technologies, such as the 10X Chromium system, use a microfluidic system and a set of specially designed 3 barcodes (aka UIDs) to tag short DNA reads which were originally sourced from the same long fragment of DNA; subsequently, these specially barcoded reads are sequenced on standard short read platforms. This approach results in interesting compromises. Each long fragment of DNA is covered only sparsely by short reads, no information about the relative ordering of reads from the same fragment is preserved, and typically each 3 barcode matches reads from 2-20 long fragments of DNA. However, compared to long read platforms like those produced by Pacific Biosciences and Oxford Nanopore the cost per base to sequence is far lower, far less input DNA is required, and the per base error rate is that of Illumina short-reads.\n\nThe use of Linked-Reads presents a new set of algorithmic challenges. In this paper, we formally describe one particular issue common to all applications of Linked-Read technology: the deconvolution of reads with a single 3 barcode into clusters that correspond to a single long fragment of DNA. We introduce Minerva, A graph-based algorithm that approximately solves the barcode deconvolution problem for metagenomic data (where reference genomes may be incomplete or unavailable). Additionally, we develop two demonstrations where the deconvolution of barcoded reads improves downstream results: improving the specificity of taxonomic assignments, and by improving clustering of related sequences. To the best of our knowledge, we are the first to address the problem of barcode deconvolution in metagenomics.

bioinformatics↗