Bioinformatics advances最新文献

Motivation: Charged amino acid residues on the spike protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have been shown to influence its binding to different cell surface receptors, its non-specific electrostatic interactions with the environment, and its structural stability and conformation. It is therefore important to obtain a good understanding of amino acid mutations that affect the total charge on the spike protein which have arisen across different SARS-CoV-2 lineages during the course of the virus' evolution.

Results: We analyse the change in the number of ionizable amino acids and the corresponding total charge on the spike proteins of almost 2200 SARS-CoV-2 lineages that have emerged over the span of the pandemic. Our results show that the previously observed trend toward an increase in the positive charge on the spike protein of SARS-CoV-2 variants of concern has essentially stopped with the emergence of the early omicron variants. Furthermore, recently emerged lineages show a greater diversity in terms of their composition of ionizable amino acids. We also demonstrate that the patterns of change in the number of ionizable amino acids on the spike protein are characteristic of related lineages within the broader clade division of the SARS-CoV-2 phylogenetic tree. Due to the ubiquity of electrostatic interactions in the biological environment, our findings are relevant for a broad range of studies dealing with the structural stability of SARS-CoV-2 and its interactions with the environment.

Availability and implementation: The data underlying the article are available in the Supplementary material.

Background: Software for labeling biological sequences typically produces a theory-based statistic for each match (the E-value) that indicates the likelihood of seeing that match's score by chance. E-values accurately predict false match rate for comparisons of random (shuffled) sequences, and thus provide a reasoned mechanism for setting score thresholds that enable high sensitivity with low expected false match rate. This threshold-setting strategy is challenged by real biological sequences, which contain regions of local repetition and low sequence complexity that cause excess matches between non-homologous sequences. Knowing this, tool developers often develop benchmarks that use realistic-seeming decoy sequences to explore empirical tradeoffs between sensitivity and false match rate. A recent trend has been to employ reversed biological sequences as realistic decoys, because these preserve the distribution of letters and the existence of local repeats, while disrupting the original sequence's functional properties. However, we and others have observed that sequences appear to produce high scoring alignments to their reversals with surprising frequency, leading to overstatement of false match risk that may negatively affect downstream analysis.

Results: We demonstrate that an alignment between a sequence S and its (possibly mutated) reversal tends to produce higher scores than alignment between truly unrelated sequences, even when S is a shuffled string with no notable repetitive or low-complexity regions. This phenomenon is due to the unintuitive fact that (even randomly shuffled) sequences contain palindromes that are on average longer than the longest common substrings (LCS) shared between permuted variants of the same sequence. Though the expected palindrome length is only slightly larger than the expected LCS, the distribution of alignment scores involving reversed sequences is strongly right-shifted, leading to greatly increased frequency of high-scoring alignments to reversed sequences.

Impact: Overestimates of false match risk can motivate unnecessarily high score thresholds, leading to potentially reduced true match sensitivity. Also, when tool sensitivity is only reported up to the score of the first matched decoy sequence, a large decoy set consisting of reversed sequences can obscure sensitivity differences between tools. As a result of these observations, we advise that reversed biological sequences be used as decoys only when care is taken to remove positive matches in the original (un-reversed) sequences, or when overstatement of false labeling is not a concern. Though the primary focus of the analysis is on sequence annotation, we also demonstrate that the prevalence of internal palindromes may lead to an overstatement of the rate of false labels in protein identification with mass spectrometry.

Motivation: Annotating cell types is a challenging yet essential task in analyzing single-cell RNA sequencing data. However, due to the lack of a gold standard, it is difficult to evaluate the algorithms fairly and an overfitting algorithm may be favored in benchmarks. To address this challenge, we developed a deep learning-based single-cell type prediction tool that assigns the cell type to 265 different cell types for humans, based on data from approximately five million cells.

Results: We achieved a median area under the ROC curve (AUC) of 0.93 when evaluated across datasets. We found that inconsistent labeling in the existing database generated by different labs contributed to the mistakes of the model. Therefore, we used cell ontology to correct the annotations and retrained the model, which resulted in 0.971 median AUC. Our study reveals a limiting factor of the accuracy one may achieve with the current database annotation and points to the solutions towards an algorithm-based correction of the gold standard for future automated cell annotation approaches.

Availability and implementation: The code is available at: https://github.com/SherrySDong/Hierarchical-Correction-Improves-Automated-Single-cell-Type-Annotation. Data used in this study are listed in Supplementary Table S1 and are retrievable at the CZI database.

Summary: SHapley Additive exPlanations (SHAP) is a widely used method for model interpretation. However, its full potential often remains untapped due to the absence of dedicated software tools. In response, ExplaineR, an R package to facilitate interpretation of binary classification and regression models based on clustering functionality for SHAP analysis is introduced here. It additionally offers user-interactive elements in visualizations for evaluating model performance, fairness analysis, decision-curve analysis, and a diverse range of SHAP plots. It facilitates in-depth post-prediction analysis of models, enabling users to pinpoint potentially significant patterns in SHAP plots and subsequently trace them back to instances through SHAP clustering. This functionality is particularly valuable for identifying patient subgroups in clinical cohorts, thus enhancing its role as a robust profiling tool. ExplaineR empowers users to generate comprehensive reports on machine learning outcomes, ensuring consistent and thorough documentation of model performance and interpretations.

Availability and implementation: ExplaineR 1.0.0 is available on GitHub (https://persimune.github.io/explainer/) and CRAN (https://cran.r-project.org/web/packages/explainer/index.html).

Motivation: Predicting anticancer treatment response from baseline genomic data is a critical obstacle in personalized medicine. Machine learning methods are commonly used for predicting drug response from gene expression data. In the process of constructing these machine learning models, one of the most significant challenges is identifying appropriate features among a massive number of genes.

Results: In this study, we utilize features (genes) extracted using the text-mining of scientific literatures. Using two independent cancer pharmacogenomic datasets, we demonstrate that text-mining-based features outperform traditional feature selection techniques in machine learning tasks. In addition, our analysis reveals that text-mining feature-based machine learning models trained on in vitro data also perform well when predicting the response of in vivo cancer models. Our results demonstrate that text-mining-based feature selection is an easy to implement approach that is suitable for building machine learning models for anticancer drug response prediction.

Availability and implementation: https://github.com/merlab/text_features.

Motivation: Cell-type deconvolution methods aim to infer cell composition from bulk transcriptomic data. The proliferation of developed methods coupled with inconsistent results obtained in many cases, highlights the pressing need for guidance in the selection of appropriate methods. Additionally, the growing accessibility of single-cell RNA sequencing datasets, often accompanied by bulk expression from related samples enable the benchmark of existing methods.

Results: In this study, we conduct a comprehensive assessment of 31 methods, utilizing single-cell RNA-sequencing data from diverse human and mouse tissues. Employing various simulation scenarios, we reveal the efficacy of regression-based deconvolution methods, highlighting their sensitivity to reference choices. We investigate the impact of bulk-reference differences, incorporating variables such as sample, study and technology. We provide validation using a gold standard dataset from mononuclear cells and suggest a consensus prediction of proportions when ground truth is not available. We validated the consensus method on data from the stomach and studied its spillover effect. Importantly, we propose the use of the critical assessment of transcriptomic deconvolution (CATD) pipeline which encompasses functionalities for generating references and pseudo-bulks and running implemented deconvolution methods. CATD streamlines simultaneous deconvolution of numerous bulk samples, providing a practical solution for speeding up the evaluation of newly developed methods.

Availability and implementation: https://github.com/Papatheodorou-Group/CATD_snakemake.

We present dingo, a Python package that supports a variety of methods to sample from the flux space of metabolic models, based on state-of-the-art random walks and rounding methods. For uniform sampling, dingo's sampling methods provide significant speed-ups and outperform existing software. Indicatively, dingo can sample from the flux space of the largest metabolic model up to now (Recon3D) in less than a day using a personal computer, under several statistical guarantees; this computation is out of reach for other similar software. In addition, dingo supports common analysis methods, such as flux balance analysis and flux variability analysis, and visualization components. dingo contributes to the arsenal of tools in metabolic modelling by enabling flux sampling in high dimensions (in the order of thousands).

Availability and implementation: The dingo Python library is available in GitHub at https://github.com/GeomScale/dingo and the data underlying this article are available in https://doi.org/10.5281/zenodo.10423335.