AdaSampling for Positive-Unlabeled and Label Noise Learning With Bioinformatics Applications

Predicting host range expansion in parasitic mites using a global mammalian-acarine dataset

Multi-host parasites pose greater health risks to wildlife, livestock, and humans than single-host parasites, yet our understanding of how ecological and biological factors influence a parasite's host range remains limited. Here, we assemble the largest and most complete dataset on permanently parasitic mammalian mites and build a predictive model assessing the probability of single-host parasites to become multi-hosts, while accounting for potentially unobserved host-parasite links and class imbalance. This model identifies statistically significant predictors related to parasites, hosts, climate, and habitat disturbance. The most important predictors include the parasite's contact level with the host immune system and two variables characterizing host phylogenetic similarity and spatial co-distribution. Our model reveals an overrepresentation of mites associated with Rodentia (rodents), Chiroptera (bats), and Carnivora in the multi-host risk group. This highlights both the potential vulnerability of these hosts to parasitic infestations and the risk of serving as reservoirs of parasites for new hosts. In addition, we find independent macroevolutionary evidence that supports our prediction of several single-host species of Notoedres, the bat skin parasites, to be in the multi-host risk group, demonstrating the forecasting potential of our model.

AI-guided pipeline for protein-protein interaction drug discovery identifies a SARS-CoV-2 inhibitor

Protein-protein interactions (PPIs) offer great opportunities to expand the druggable proteome and therapeutically tackle various diseases, but remain challenging targets for drug discovery. Here, we provide a comprehensive pipeline that combines experimental and computational tools to identify and validate PPI targets and perform early-stage drug discovery. We have developed a machine learning approach that prioritizes interactions by analyzing quantitative data from binary PPI assays or AlphaFold-Multimer predictions. Using the quantitative assay LuTHy together with our machine learning algorithm, we identified high-confidence interactions among SARS-CoV-2 proteins for which we predicted three-dimensional structures using AlphaFold-Multimer. We employed VirtualFlow to target the contact interface of the NSP10-NSP16 SARS-CoV-2 methyltransferase complex by ultra-large virtual drug screening. Thereby, we identified a compound that binds to NSP10 and inhibits its interaction with NSP16, while also disrupting the methyltransferase activity of the complex, and SARS-CoV-2 replication. Overall, this pipeline will help to prioritize PPI targets to accelerate the discovery of early-stage drug candidates targeting protein complexes and pathways.

SnapKin: a snapshot deep learning ensemble for kinase-substrate prediction from phosphoproteomics data

A major challenge in mass spectrometry-based phosphoproteomics lies in identifying the substrates of kinases, as currently only a small fraction of substrates identified can be confidently linked with a known kinase. Machine learning techniques are promising approaches for leveraging large-scale phosphoproteomics data to computationally predict substrates of kinases. However, the small number of experimentally validated kinase substrates (true positive) and the high data noise in many phosphoproteomics datasets together limit their applicability and utility. Here, we aim to develop advanced kinase-substrate prediction methods to address these challenges. Using a collection of seven large phosphoproteomics datasets, and both traditional and deep learning models, we first demonstrate that a 'pseudo-positive' learning strategy for alleviating small sample size is effective at improving model predictive performance. We next show that a data resampling-based ensemble learning strategy is useful for improving model stability while further enhancing prediction. Lastly, we introduce an ensemble deep learning model ('SnapKin') by incorporating the above two learning strategies into a 'snapshot' ensemble learning algorithm. We propose SnapKin, an ensemble deep learning method, for predicting substrates of kinases from large-scale phosphoproteomics data. We demonstrate that SnapKin consistently outperforms existing methods in kinase-substrate prediction. SnapKin is freely available at https://github.com/PYangLab/SnapKin.

Automatic quality control of single-cell and single-nucleus RNA-seq using valiDrops

Single-cell and single-nucleus RNA-sequencing (sxRNA-seq) measures gene expression in individual cells or nuclei enabling comprehensive characterization of cell types and states. However, isolation of cells or nuclei for sxRNA-seq releases contaminating RNA, which can distort biological signals, through, for example, cell damage and transcript leakage. Thus, identifying barcodes containing high-quality cells or nuclei is a critical analytical step in the processing of sxRNA-seq data. Here, we present valiDrops, an automated method to identify high-quality barcodes and flag dead cells. In valiDrops, barcodes are initially filtered using data-adaptive thresholding on community-standard quality metrics, and subsequently, valiDrops uses a novel clustering-based approach to identify barcodes with distinct biological signals. We benchmark valiDrops and show that biological signals from cell types and states are more distinct, easier to separate and more consistent after filtering by valiDrops compared to existing tools. Finally, we show that valiDrops can predict and flag dead cells with high accuracy. This novel classifier can further improve data quality or be used to identify dead cells to interrogate the biology of cell death. Thus, valiDrops is an effective and easy-to-use method to improve data quality and biological interpretation. Our method is openly available as an R package at www.github.com/madsen-lab/valiDrops.

AI-guided pipeline for protein-protein interaction drug discovery identifies a SARS-CoV-2 inhibitor

Protein-protein interactions (PPIs) offer great opportunities to expand the druggable proteome and therapeutically tackle various diseases, but remain challenging targets for drug discovery. Here, we provide a comprehensive pipeline that combines experimental and computational tools to identify and validate PPI targets and perform early-stage drug discovery. We have developed a machine learning approach that prioritizes interactions by analyzing quantitative data from binary PPI assays and AlphaFold-Multimer predictions. Using the quantitative assay LuTHy together with our machine learning algorithm, we identified high-confidence interactions among SARS-CoV-2 proteins for which we predicted three-dimensional structures using AlphaFold Multimer. We employed VirtualFlow to target the contact interface of the NSP10-NSP16 SARS-CoV-2 methyltransferase complex by ultra-large virtual drug screening. Thereby, we identified a compound that binds to NSP10 and inhibits its interaction with NSP16, while also disrupting the methyltransferase activity of the complex, and SARS-CoV-2 replication. Overall, this pipeline will help to prioritize PPI targets to accelerate the discovery of early-stage drug candidates targeting protein complexes and pathways.

A multi-omics integrative analysis based on CRISPR screens re-defines the pluripotency regulatory network in ESCs

A comprehensive and precise definition of the pluripotency gene regulatory network (PGRN) is crucial for clarifying the regulatory mechanisms in embryonic stem cells (ESCs). Here, after a CRISPR/Cas9-based functional genomics screen and integrative analysis with other functional genomes, transcriptomes, proteomes and epigenome data, an expanded pluripotency-associated gene set is obtained, and a new PGRN with nine sub-classes is constructed. By integrating the DNA binding, epigenetic modification, chromatin conformation, and RNA expression profiles, the PGRN is resolved to six functionally independent transcriptional modules (CORE, MYC, PAF, PRC, PCGF and TBX). Spatiotemporal transcriptomics reveal activated CORE/MYC/PAF module activity and repressed PRC/PCGF/TBX module activity in both mouse ESCs (mESCs) and pluripotent cells of early embryos. Moreover, this module activity pattern is found to be shared by human ESCs (hESCs) and cancers. Thus, our results provide novel insights into elucidating the molecular basis of ESC pluripotency.

MGRCDA: Metagraph Recommendation Method for Predicting CircRNA-Disease Association

Clinical evidence began to accumulate, suggesting that circRNAs can be novel therapeutic targets for various diseases and play a critical role in human health. However, limited by the complex mechanism of circRNA, it is difficult to quickly and large-scale explore the relationship between disease and circRNA in the wet-lab experiment. In this work, we design a new computational model MGRCDA on account of the metagraph recommendation theory to predict the potential circRNA-disease associations. Specifically, we first regard the circRNA-disease association prediction problem as the system recommendation problem, and design a series of metagraphs according to the heterogeneous biological networks; then extract the semantic information of the disease and the Gaussian interaction profile kernel (GIPK) similarity of circRNA and disease as network attributes; finally, the iterative search of the metagraph recommendation algorithm is used to calculate the scores of the circRNA-disease pair. On the gold standard dataset circR2Disease, MGRCDA achieved a prediction accuracy of 92.49% with an area under the ROC curve of 0.9298, which is significantly higher than other state-of-the-art models. Furthermore, among the top 30 disease-related circRNAs recommended by the model, 25 have been verified by the latest published literature. The experimental results prove that MGRCDA is feasible and efficient, and it can recommend reliable candidates to further wet-lab experiment and reduce the scope of the experiment.

A hybrid approach for noise reduction-based optimal classifier using genetic algorithm: A case study in plant disease prediction

Plant diseases can cause significant losses to agricultural productivity; therefore, their early prediction is much needed. So far, many machine learning-based plant disease prediction models have been recommended, but these models face a problem of noisy class label dataset that degrades the performance. Noisy class label dataset results from the improper assignment of positive class labels into negative class data samples or vice versa. Hence, a precise and noise-free plant disease model is required for a better prediction. The current study proposes noise reduction-based hybridized classifiers for plant disease prediction. One tomato and four soybean disease datasets have been selected to conduct the proposed research. The Adaptive Sampling-based Class Label Noise Reduction (AS-CLNR) method has been used along with the Support Vector Machine (SVM) approach for noise reduction. The noise-minimized datasets have been fed into the Extreme Learning Machine (ELM), Decision Tree (DT), and Random Forest (RF) classifiers whose parameters are optimized using Genetic Algorithm (GA) for developing plant disease prediction models. The performances of all these models viz. Hybrid SVM-GA-ELM, Hybrid SVM-GA-DT, and Hybrid SVM-GA-RF have been evaluated using Accuracy, Area under ROC Curve, and F1-Score metrics. Further, these classifiers have been ranked using the statistical Friedman Test in which the Hybrid SVM-GA-RF classifier performed the best. Lastly, the Nemenyi test has also been performed to find out if significant differences exist between various classifiers or not. It was found that 33.33% of the total pairs of hybrid classifiers show a remarkably different performance from one another.

PLUS: Predicting cancer metastasis potential based on positive and unlabeled learning

Metastatic cancer accounts for over 90% of all cancer deaths, and evaluations of metastasis potential are vital for minimizing the metastasis-associated mortality and achieving optimal clinical decision-making. Computational assessment of metastasis potential based on large-scale transcriptomic cancer data is challenging because metastasis events are not always clinically detectable. The under-diagnosis of metastasis events results in biased classification labels, and classification tools using biased labels may lead to inaccurate estimations of metastasis potential. This issue is further complicated by the unknown metastasis prevalence at the population level, the small number of confirmed metastasis cases, and the high dimensionality of the candidate molecular features. Our proposed algorithm, called Positive and unlabeled Learning from Unbalanced cases and Sparse structures (PLUS), is the first to use a positive and unlabeled learning framework to account for the under-detection of metastasis events in building a classifier. PLUS is specifically tailored for studying metastasis that deals with the unbalanced instance allocation as well as unknown metastasis prevalence, which are not considered by other methods. PLUS achieves superior performance on synthetic datasets compared with other state-of-the-art methods. Application of PLUS to The Cancer Genome Atlas Pan-Cancer gene expression data generated metastasis potential predictions that show good agreement with the clinical follow-up data, in addition to predictive genes that have been validated by independent single-cell RNA-sequencing datasets.

Metastasis is the major cause of cancer-related deaths, and evaluations of metastasis risk are essential for tailored treatment of cancer patients. Existing methods often build a classifier using the clinical metastasis diagnoses as binary responses or detect genomic features significantly associated with metastasis-related survival outcomes. However, these methods tend to identify genomic predictors that have little consistency across different cancer types. Thus, there is an urgent need for a powerful tool to characterize the cancer metastasis potential applicable across a wide span of cancer types. Computational assessment of metastasis potential based on large-scale transcriptomic cancer data is challenging because metastasis events are not always clinically detectable, which results in biased estimations of metastasis potential. Our proposed algorithm, called PLUS, considers patients with metastasis diagnosis as positive instances and the remainder as unlabeled instances, meaning they are either metastatic or non-metastatic. Such a classifier given by PLUS rendered concordance between the predicted cancer metastasis and observed metastasis survival outcomes in the follow-up data for almost all cancer types considered. The selected genes were found to perform functions consistent with experimental research findings and are capable of clustering the single cells based on their levels of metastasis potential.

Co-evolution based machine-learning for predicting functional interactions between human genes

Over the next decade, more than a million eukaryotic species are expected to be fully sequenced. This has the potential to improve our understanding of genotype and phenotype crosstalk, gene function and interactions, and answer evolutionary questions. Here, we develop a machine-learning approach for utilizing phylogenetic profiles across 1154 eukaryotic species. This method integrates co-evolution across eukaryotic clades to predict functional interactions between human genes and the context for these interactions. We benchmark our approach showing a 14% performance increase (auROC) compared to previous methods. Using this approach, we predict functional annotations for less studied genes. We focus on DNA repair and verify that 9 of the top 50 predicted genes have been identified elsewhere, with others previously prioritized by high-throughput screens. Overall, our approach enables better annotation of function and functional interactions and facilitates the understanding of evolutionary processes underlying co-evolution. The manuscript is accompanied by a webserver available at: https://mlpp.cs.huji.ac.il.

With the rise in number of eukaryotic species being fully sequenced, large scale phylogenetic profiling can give insights on gene function, Here, the authors describe a machine-learning approach that integrates co-evolution across eukaryotic clades to predict gene function and functional interactions among human genes.

Protocol for the processing and downstream analysis of phosphoproteomic data with PhosR

Analysis of phosphoproteomic data requires advanced computational methodologies. To this end, we developed PhosR, a set of tools and methodologies implemented in R to allow the comprehensive analysis of phosphoproteomic data. PhosR enables processing steps such as imputation, normalization, and functional analysis such as kinase activity inference and signalome construction. Together, PhosR facilitates interpretation and discovery from large-scale phosphoproteomic data sets.

For complete details on the use and execution of this protocol, please refer to Kim et al. (2021).

•

This protocol describes how to run and interpret the results of PhosR

•

PhosR performs filtering, imputation, and normalization of phosphoproteomic data

•

PhosR enables kinase-substrate predictions and signalome construction

•

The step-by-step protocol provides a comprehensive introduction to phospho-data analysis

Semi-Supervised Learning Using Hierarchical Mixture Models: Gene Essentiality Case Study

Integrating gene-level data is useful for predicting the role of genes in biological processes. This problem has typically focused on supervised classification, which requires large training sets of positive and negative examples. However, training data sets that are too small for supervised approaches can still provide valuable information. We describe a hierarchical mixture model that uses limited positively labeled gene training data for semi-supervised learning. We focus on the problem of predicting essential genes, where a gene is required for the survival of an organism under particular conditions. We applied cross-validation and found that the inclusion of positively labeled samples in a semi-supervised learning framework with the hierarchical mixture model improves the detection of essential genes compared to unsupervised, supervised, and other semi-supervised approaches. There was also improved prediction performance when genes are incorrectly assumed to be non-essential. Our comparisons indicate that the incorporation of even small amounts of existing knowledge improves the accuracy of prediction and decreases variability in predictions. Although we focused on gene essentiality, the hierarchical mixture model and semi-supervised framework is standard for problems focused on prediction of genes or other features, with multiple data types characterizing the feature, and a small set of positive labels.

PhosR enables processing and functional analysis of phosphoproteomic data

Mass spectrometry (MS)-based phosphoproteomics has revolutionized our ability to profile phosphorylation-based signaling in cells and tissues on a global scale. To infer the action of kinases and signaling pathways in phosphoproteomic experiments, we present PhosR, a set of tools and methodologies implemented in a suite of R packages facilitating comprehensive analysis of phosphoproteomic data. By applying PhosR to both published and new phosphoproteomic datasets, we demonstrate capabilities in data imputation and normalization by using a set of "stably phosphorylated sites" and in functional analysis for inferring active kinases and signaling pathways. In particular, we introduce a "signalome" construction method for identifying a collection of signaling modules to summarize and visualize the interaction of kinases and their collective actions on signal transduction. Together, our data and findings demonstrate the utility of PhosR in processing and generating biological knowledge from MS-based phosphoproteomic data.

Transcriptional network dynamics during the progression of pluripotency revealed by integrative statistical learning

The developmental potential of cells, termed pluripotency, is highly dynamic and progresses through a continuum of naive, formative and primed states. Pluripotency progression of mouse embryonic stem cells (ESCs) from naive to formative and primed state is governed by transcription factors (TFs) and their target genes. Genomic techniques have uncovered a multitude of TF binding sites in ESCs, yet a major challenge lies in identifying target genes from functional binding sites and reconstructing dynamic transcriptional networks underlying pluripotency progression. Here, we integrated time-resolved ‘trans-omic’ datasets together with TF binding profiles and chromatin conformation data to identify target genes of a panel of TFs. Our analyses revealed that naive TF target genes are more likely to be TFs themselves than those of formative TFs, suggesting denser hierarchies among naive TFs. We also discovered that formative TF target genes are marked by permissive epigenomic signatures in the naive state, indicating that they are poised for expression prior to the initiation of pluripotency transition to the formative state. Finally, our reconstructed transcriptional networks pinpointed the precise timing from naive to formative pluripotency progression and enabled the spatiotemporal mapping of differentiating ESCs to their in vivo counterparts in developing embryos.

Probing lncRNA-Protein Interactions: Data Repositories, Models, and Algorithms

Identifying lncRNA-protein interactions (LPIs) is vital to understanding various key biological processes. Wet experiments found a few LPIs, but experimental methods are costly and time-consuming. Therefore, computational methods are increasingly exploited to capture LPI candidates. We introduced relevant data repositories, focused on two types of LPI prediction models: network-based methods and machine learning-based methods. Machine learning-based methods contain matrix factorization-based techniques and ensemble learning-based techniques. To detect the performance of computational methods, we compared parts of LPI prediction models on Leave-One-Out cross-validation (LOOCV) and fivefold cross-validation. The results show that SFPEL-LPI obtained the best performance of AUC. Although computational models have efficiently unraveled some LPI candidates, there are many limitations involved. We discussed future directions to further boost LPI predictive performance.

scReClassify: post hoc cell type classification of single-cell rNA-seq data

Background

Single-cell RNA-sequencing (scRNA-seq) is a fast emerging technology allowing global transcriptome profiling on the single cell level. Cell type identification from scRNA-seq data is a critical task in a variety of research such as developmental biology, cell reprogramming, and cancers. Typically, cell type identification relies on human inspection using a combination of prior biological knowledge (e.g. marker genes and morphology) and computational techniques (e.g. PCA and clustering). Due to the incompleteness of our current knowledge and the subjectivity involved in this process, a small amount of cells may be subject to mislabelling.

Results

Here, we propose a semi-supervised learning framework, named scReClassify, for ‘post hoc’ cell type identification from scRNA-seq datasets. Starting from an initial cell type annotation with potentially mislabelled cells, scReClassify first performs dimension reduction using PCA and next applies a semi-supervised learning method to learn and subsequently reclassify cells that are likely mislabelled initially to the most probable cell types. By using both simulated and real-world experimental datasets that profiled various tissues and biological systems, we demonstrate that scReClassify is able to accurately identify and reclassify misclassified cells to their correct cell types.

Conclusions

scReClassify can be used for scRNA-seq data as a post hoc cell type classification tool to fine-tune cell type annotations generated by any cell type classification procedure. It is implemented as an R package and is freely available from https://github.com/SydneyBioX/scReClassify