Research in Computational Molecular Biology: 26th Annual International Conference, RECOMB 2022, San Diego, CA, USA, May 22–25, 2022, Proceedings (Lecture Notes in Bioinformatics)

This book constitutes the proceedings of the 26th Annual Conference on Research in Computational Molecular Biology, RECOMB 2022, held in San Diego, CA, USA in May 2022. The 17 regular and 23 short papers presented were carefully reviewed and selected from 188 submissions. The papers report on original research in all areas of computational molecular biology and bioinformatics. Preface To Benny, to the RECOMB Community in Memory of Benny Organization Contents Extended Abstracts Unsupervised Integration of Single-Cell Multi-omics Datasets with Disproportionate Cell-Type Representation 1 Introduction 2 Method 2.1 Unbalanced Optimal Transport of SCOTv2 2.2 Extending SCOTv2 for Multi-domain Alignment 2.3 Embedding with the Coupling Matrix 2.4 Heuristic Process for Self-tuning Hyperparameters 3 Experimental Setup 3.1 Datasets 3.2 Evaluation Metrics and Baseline Methods 4 Results 4.1 SCOTv2 Gives High-Quality Alignments Consistently Across All Single-Datasets 4.2 Hyperparameter Self-tuning Aligns Well Without Depending on Orthogonal Correspondence Information 4.3 SCOTv2 Scales Well with Increasing Number of Samples 5 Discussion References Semi-supervised Single-Cell Cross-modality Translation Using Polarbear 1 Introduction 1.1 Related Work 2 Methods 2.1 Polarbear Model 2.2 Hyperparameter Tuning 2.3 Performance Measures 2.4 Single-Cell Data Pre-processing 2.5 Cluster-Level Analysis 3 Results 3.1 Polarbear Accurately Translates Between Single-Cell Data Domains 3.2 Polarbear Generalizes to New Cell Types 3.3 Polarbear Can Match Corresponding Cells Across Modalities 4 Discussion References Transcription Factor-Centric Approach to Identify Non-recurring Putative Regulatory Drivers in Cancer 1 Introduction 2 Data and Methods 2.1 ICGC Simple Somatic Mutations and Gene Expression Data 2.2 Promoter and Enhancer Data 2.3 Defining the Effects of Mutations on TF Binding, and the Significance of These Effects 2.4 Analytical and Simulation-Based Approaches to Compute the Significance of Mutation Effects on TF Binding 2.5 Integrating Results Across All Regulatory Regions of a Gene 3 Results 3.1 Integrated Analysis Across Regulatory Regions Identifies 54 Genes with Significant TF Binding Changes Due to Mutations in Regulatory DNA 3.2 Genes with Significant Mutations in Their Regulatory Regions Show Large Expression Differences in Mutated Versus Non-mutated Samples 4 Discussion 5 Acknowledgements, Code Availability, and Supplemental Materials References DeepMinimizer: A Differentiable Framework for Optimizing Sequence-Specific Minimizer Schemes 1 Introduction 2 Related Work 3 Methods 3.1 Background 3.2 Search Space Reparameterization 3.3 Proxy Objective 3.4 Specification of TemplateNet 3.5 Specification of the Divergence Measure 4 Results 5 Conclusion A Proof of Proposition 1 B Other Empirical Results References MetaCoAG: Binning Metagenomic Contigs via Composition, Coverage and Assembly Graphs 1 Introduction 2 Methods 2.1 Step 0: Assemble Reads into Contigs and Construct the Assembly Graph 2.2 Step 1: Identify Contigs with Single-Copy Marker Genes 2.3 Step 2: Order Single-Copy Marker Genes and Estimate the Number of Initial Bins 2.4 Step 3: Bin Contigs with Single-copy Marker Genes 2.5 Step 4: Bin Remaining Contigs Using Label Propagation 3 Experimental Setup 3.1 Datasets and Tools 3.2 Evaluation Metrics 4 Results and Discussion 4.1 Benchmarks Using SimHC+ Dataset 4.2 Benchmarks Using Real Datasets 5 Discussion and Conclusion A Appendix References A Fast, Provably Accurate Approximation Algorithm for Sparse Principal Component Analysis Reveals Human Genetic Variation Across the World 1 Introduction 1.1 Our Contributions 1.2 Prior Work 2 Materials and Methods 2.1 The ThreSPCA Algorithm 2.2 Data 2.3 Experiments 3 Results 3.1 ThreSPCA Reveals Genetic Diversity Across the World 3.2 Interpretability of ThreSPCA Informed Variants 3.3 Comparing ThreSPCA to State-of-the-Art 4 Discussion Appendix 1.A SPCA via Thresholding: Discussions and Proofs Appendix 1.B Additional Experiments Appendix 1.B.1 Simulated Studies Appendix 1.B.2 Experiments on 1KG data Appendix 1.B.3 Comparing ThreSPCA with the State-of-the-Art References Gene Set Priorization Guided by Regulatory Networks with p-values through Kernel Mixed Model 1 Introduction 2 Method 2.1 Background 2.2 Method 3 Simulation Experiments 3.1 Competing Methods 3.2 General Data Generation Process 3.3 Results 4 Study of Transcriptome Association of Alzheimer's Disease 5 Conclusion A Additional Simulation Experiments B Covaraite Regressing References Real-Valued Group Testing for Quantitative Molecular Assays 1 Introduction 1.1 Problem Statement and Contribution 2 Methods 2.1 Notation 2.2 Overview of the Matrix Design and Decoding Algorithms 2.3 Constructing Matrices for Real-Valued Group Testing 3 Results 3.1 Comparison of Matrix Properties with Existing Approaches 3.2 Effectiveness on Simulated Data 3.3 Effectiveness in Wet Lab 4 Conclusion References On the Effect of Intralocus Recombination on Triplet-Based Species Tree Estimation 1 Introduction 1.1 Key Definitions 1.2 Inference Methods 1.3 Multispecies Coalescent with Recombination 1.4 Estimating Sequence Distances 2 Inconsistency of R^* 2.1 Statement and Overview 2.2 Key Lemmas 2.3 Proof of Theorem 1 3 Simulation Study 4 Discussion References QT-GILD: Quartet Based Gene Tree Imputation Using Deep Learning Improves Phylogenomic Analyses Despite Missing Data 1 Introduction 2 Quartet Imputation Problem 2.1 Problem Definition 3 Experimental Study 3.1 Datasets 3.2 Generating Incomplete Gene Trees 3.3 Species Tree Estimation Methods 3.4 Measurements 4 Results and Discussion 4.1 Results on 15-Taxon Dataset 4.2 Results on 37-Taxon Mammalian Simulated Dataset 4.3 Results on Biological Dataset 4.4 Running Time 5 Conclusions References Safety and Completeness in Flow Decompositions for RNA Assembly 1 Introduction 1.1 Safety Framework for Addressing Multiple Solutions 1.2 Safety in Flow Decomposition for RNA Assembly 1.3 Our Results 2 Preliminaries and Notations 3 Characterization of Safe and Complete Paths 4 Simple Verification and Enumeration Algorithms 5 Experimental Evaluation 5.1 Datasets 5.2 Evaluation Metrics 5.3 Implementation and Environment Details 5.4 Results 6 Conclusion References NetMix2: Unifying Network Propagation and Altered Subnetworks 1 Introduction 2 Methods 2.1 Altered Subnetwork Problem 2.2 Network Propagation and the Propagation Family 2.3 NetMix2 2.4 Scores-Only and Network-Only Baselines 3 Results 3.1 Somatic Mutations in Cancer 4 Discussion References Multi-modal Genotype and Phenotype Mutual Learning to Enhance Single-Modal Input Based Longitudinal Outcome Prediction 1 Introduction 2 Related Work 3 Proposed Method 3.1 Problem Formulation 3.2 Notation 3.3 Longitudinal Predictive Model 4 Experiments 4.1 Experimental Setup 4.2 Experimental Results 5 Conclusion References Fast, Flexible, and Exact Minimum Flow Decompositions via ILP 1 Introduction 1.1 Minimum Flow Decomposition in Multiassembly 1.2 Limitations of Current ILP Solutions 1.3 Our Contributions 2 Preliminaries 3 ILP Formulations 3.1 Minimum Flow Decomposition 3.2 Subpath Constraints 3.3 Inexact Flow 4 Experiments 5 Conclusions References Co-linear Chaining with Overlaps and Gap Costs 1 Introduction 2 Concepts and Definitions 2.1 Co-linear Chaining Problem with Overlap and Gap Costs 2.2 Anchored Edit Distance 2.3 Graph Representation of Alignment 3 Our Algorithms 4 Proof of Equivalence 4.1 Details of Lemma 2 Proof 5 Implementation 6 Evaluation References The Complexity of Approximate Pattern Matching on de Bruijn Graphs 1 Introduction 1.1 Technical Background and Our Results 2 NP-Completeness of Problem 1 on de Bruijn Graphs 2.1 Reduction 3 Hardness for Problem 2 on de Bruijn Graphs 3.1 Proof of Correctness 4 Discussion References ProTranslator: Zero-Shot Protein Function Prediction Using Textual Description 1 Introduction 2 Methods 2.1 Problem Definition 2.2 Embedding GO Functions Based on the Textual Description 2.3 Embedding Proteins Based on Sequence, Description and Network 2.4 Protein Function Prediction Based on GO Embeddings and Protein Embeddings 2.5 Annotate Novel Functions, Sparse Functions and Gene Sets to Pathways 2.6 Text Generation by the Protein Sequence Features 3 Experimental Setup 3.1 Calculating Similarities Between GO Functions 3.2 Datasets and Evaluation 3.3 Comparison Approaches 4 Results 4.1 Gene Ontology Term Description Similarity Reflects Function Annotation Similarity 4.2 ProTranslator Enables Protein Function Prediction in the Zero-Shot Setting 4.3 ProTranslator Obtains Substantial Improvement in the Few-Shot Setting 4.4 ProTranslator Annotated Genes to Pathways by Only Using the Pathway Description 4.5 ProTranslator Generates Text Description for a Gene Set 4.6 Ablation Experiment 5 Conclusion and Discussion References Short Papers Single-Cell Multi-omic Velocity Infers Dynamic and Decoupled Gene Regulation References Ultrafast and Interpretable Single-Cell 3D Genome Analysis with Fast-Higashi DiffDomain Enables Identification of Structurally Reorganized Topologically Associating Domains Joint Inference of Repeated Evolutionary Trajectories and Patterns of Clonal Exclusivity or Co-occurrence from Tumor Mutation Trees Fast and Optimal Sequence-to-Graph Alignment Guided by Seeds 1 Introduction 2 Prerequisites 2.1 Problem Statement: Alignment as Shortest Path 2.2 A Algorithm for Finding a Shortest Path 3 Seed Heuristic 3.1 Overview 3.2 Formal Definition 3.3 Trie Index 4 Evaluations 4.1 Seed Heuristic Implementation 4.2 Setting 4.3 Q1: Speedup of the Seed Heuristic 4.4 Q2: Scaling with Reference Size 4.5 Q3: Scaling with Read Length 5 Conclusion A Appendix A.1 A Algorithm A.2 Proofs A.3 Versions, Commands, Parameters for Running all Evaluated Approaches A.4 Notations References CLMB: Deep Contrastive Learning for Robust Metagenomic Binning 1 Introduction 2 Methods 2.1 Data Augmentation 2.2 Architecture of the VAE 2.3 Loss Function 2.4 Training with Contrastive Learning 2.5 Productive Model 3 Results 3.1 Datasets and Evaluation Metrics 3.2 CLMB Recovers More Near-Complete Genomes on Most Benchmarking Datasets 3.3 The Performance of CLMB Benefits from Finding the Information of Resemblance and Discrimination Within Data 3.4 The Performance of the Ensemble Binning Is Improved by Involving CLMB 3.5 The Genomes Recovered by CLMB Assist Analysis for Mother-Infant Microbiome 4 Discussions 5 Appendix A Figures B Tables C Methods C.1 Feature Calculation of TNFs and Abundance C.2 Benchmarking C.3 Data Fusion Experiment C.4 Binning of the Mother-Infant Transmission Dataset References Unsupervised Cell Functional Annotation for Single-Cell RNA-Seq References A Novel Matrix Factorization Model for Interpreting Single-Cell Gene Expression from Biologically Heterogeneous Data References Tractable and Expressive Generative Models of Genetic Variation Data Concert: Genome-Wide Prediction of Sequence Elements That Modulate DNA Replication Timing CORSID Enables de novo Identification of Transcription Regulatory Sequences and Genes in Coronaviruses References Learning Probabilistic Protein-DNA Recognition Codes from DNA-Binding Specificities Using Structural Mappings Uncertainty Quantification Using Subsampling for Assembly-Free Estimates of Genomic Distance and Phylogenetic Relationships References SOPHIE: Viral Outbreak Investigation and Transmission History Reconstruction in a Joint Phylogenetic and Network Theory Framework 1 Introduction 2 Methods 3 Results 4 Disclaimer References Identifying Systematic Variation at the Single-Cell Level by Leveraging Low-Resolution Population-Level Data Belayer: Modeling Discrete and Continuous Spatial Variation in Gene Expression from Spatially Resolved Transcriptomics Lossless Indexing with Counting de Bruijn Graphs References Uncovering Hidden Assembly Artifacts: When Unitigs are not Safe and Bidirected Graphs are not Helpful (ABSTRACT) References Mapping Single-Cell Transcriptomes to Copy Number Evolutionary Trees ImmunoTyper-SR: A Novel Computational Approach for Genotyping Immunoglobulin Heavy Chain Variable Genes Using Short Read Data References AutoComplete: Deep Learning-Based Phenotype Imputation for Large-Scale Biomedical Data Resistor: An Algorithm for Predicting Resistance Mutations Using Pareto Optimization over Multistate Protein Design and Mutational Signatures References Ultra High Diversity Factorizable Libraries for Efficient Therapeutic Discovery References Author Index