Inference Pipeline Research Articles

Deriving star cluster parameters with convolutional neural networks

Context. Convolutional neural networks (CNNs) have been proven to perform fast classification and detection on natural images and have the potential to infer astrophysical parameters on the exponentially increasing amount of sky-survey imaging data. The inference pipeline can be trained either from real human-annotated data or simulated mock observations. Until now, star cluster analysis was based on integral or individual resolved stellar photometry. This limits the amount of information that can be extracted from cluster images. Aims. We aim to develop a CNN-based algorithm capable of simultaneously deriving ages, masses, and sizes of star clusters directly from multi-band images. We also aim to demonstrate CNN capabilities on low-mass semi-resolved star clusters in a low-signal-to-noise-ratio regime. Methods. A CNN was constructed based on the deep residual network (ResNet) architecture and trained on simulated images of star clusters with various ages, masses, and sizes. To provide realistic backgrounds, M 31 star fields taken from The Panchromatic Hubble Andromeda Treasury (PHAT) survey were added to the mock cluster images. Results. The proposed CNN was verified on mock images of artificial clusters and has demonstrated high precision and no significant bias for clusters of ages ≲3 Gyr and masses between 250 and 4000 M⊙. The pipeline is end-to-end, starting from input images all the way to the inferred parameters; no hand-coded steps have to be performed: estimates of parameters are provided by the neural network in one inferential step from raw images.

On Tracking the Physicality of Wi-Fi: A Subspace Approach

Wi-Fi channel state information (CSI) has emerged as a plausible modality for sensing different human activities as a function of modulations in the wireless signal that travels between wireless devices. Until now, most research has taken a statistical approach and/or purpose-built inference pipeline. Although interesting, these approaches struggle to sustain sensing performances beyond experimental conditions. As such, the full potential of CSI as a general-purpose sensing modality is yet to be realized. We argue that a universal approach with the well-grounded formalization is necessary to characterize the relationship between the wireless channel modulations (spatial and temporal) and human movement. To this end, we present a formalism for quantifying the changing part of the wireless signal modulated by human motion. Grounded in this formalization, we then present a new subspace tracking technique to describe the channel statistics in an interpretable way, which succinctly contains the human modulated part of the channel. We characterize the signal and noise subspaces for the case of uncontrolled human movement and show that these subspaces are dynamic. Our results demonstrate that the proposed channel statistics alone can robustly reproduce the state-of-the-art application-specific feature engineering baseline, however, across multiple usage scenarios. We expect that our universal channel statistics will yield an effective general-purpose featurization of wireless channel measurements and will uncover opportunities for applying CSI for a variety of human sensing applications in a robust way.

Identification of Microbial Dark Matter in Antarctic Environments.

Numerous studies have applied molecular techniques to understand the diversity, evolution, and ecological function of Antarctic bacteria and archaea. One common technique is sequencing of the 16S rRNA gene, which produces a nearly quantitative profile of community membership. However, the utility of this and similar approaches is limited by what is known about the evolution, physiology, and ecology of surveyed taxa. When representative genomes are available in public databases some of this information can be gleaned from genomic studies, and automated pipelines exist to carry out this task. Here the paprica metabolic inference pipeline was used to assess how well Antarctic microbial communities are represented by the available completed genomes. The NCBI’s Sequence Read Archive (SRA) was searched for Antarctic datasets that used one of the Illumina platforms to sequence the 16S rRNA gene. These data were quality controlled and denoised to identify unique reads, then analyzed with paprica to determine the degree of overlap with the closest phylogenetic neighbor with a completely sequenced genome. While some unique reads had perfect mapping to 16S rRNA genes from completed genomes, the mean percent overlap for all mapped reads was 86.6%. When samples were grouped by environment, some environments appeared more or less well represented by the available genomes. For the domain Bacteria, seawater was particularly poorly represented with a mean overlap of 80.2%, while for the domain Archaea glacial ice was particularly poorly represented with an overlap of only 48.0% for a single sample. These findings suggest that a considerable effort is needed to improve the representation of Antarctic microbes in genome sequence databases.

A unified analysis of four cosmic shear surveys

In the past few years, several independent collaborations have presented cosmological constraints from tomographic cosmic shear analyses. These analyses differ in many aspects: the datasets, the shear and photometric redshift estimation algorithms, the theory model assumptions, and the inference pipelines. To assess the robustness of the existing cosmic shear results, we present in this paper a unified analysis of four of the recent cosmic shear surveys: the Deep Lens Survey (DLS), the Canada-France-Hawaii Telescope Lensing Survey (CFHTLenS), the Science Verification data from the Dark Energy Survey (DES-SV), and the 450 deg$^{2}$ release of the Kilo-Degree Survey (KiDS-450). By using a unified pipeline, we show how the cosmological constraints are sensitive to the various details of the pipeline. We identify several analysis choices that can shift the cosmological constraints by a significant fraction of the uncertainties. For our fiducial analysis choice, considering a Gaussian covariance, conservative scale cuts, assuming no baryonic feedback contamination, identical cosmological parameter priors and intrinsic alignment treatments, we find the constraints (mean, 16% and 84% confidence intervals) on the parameter $S_{8}\equiv \sigma_{8}(\Omega_{\rm m}/0.3)^{0.5}$ to be $S_{8}=0.94_{-0.045}^{+0.046}$ (DLS), $0.66_{-0.071}^{+0.070}$ (CFHTLenS), $0.84_{-0.061}^{+0.062}$ (DES-SV) and $0.76_{-0.049}^{+0.048}$ (KiDS-450). From the goodness-of-fit and the Bayesian evidence ratio, we determine that amongst the four surveys, the two more recent surveys, DES-SV and KiDS-450, have acceptable goodness-of-fit and are consistent with each other. The combined constraints are $S_{8}=0.79^{+0.042}_{-0.041}$, which is in good agreement with the first year of DES cosmic shear results and recent CMB constraints from the Planck satellite.

Hybrid 3D/2D Convolutional Neural Network for Hemorrhage Evaluation on Head CT.

Convolutional neural networks are a powerful technology for image recognition. This study evaluates a convolutional neural network optimized for the detection and quantification of intraparenchymal, epidural/subdural, and subarachnoid hemorrhages on noncontrast CT. This study was performed in 2 phases. First, a training cohort of all NCCTs acquired at a single institution between January 1, 2017, and July 31, 2017, was used to develop and cross-validate a custom hybrid 3D/2D mask ROI-based convolutional neural network architecture for hemorrhage evaluation. Second, the trained network was applied prospectively to all NCCTs ordered from the emergency department between February 1, 2018, and February 28, 2018, in an automated inference pipeline. Hemorrhage-detection accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative predictive value were assessed for full and balanced datasets and were further stratified by hemorrhage type and size. Quantification was assessed by the Dice score coefficient and the Pearson correlation. A 10,159-examination training cohort (512,598 images; 901/8.1% hemorrhages) and an 862-examination test cohort (23,668 images; 82/12% hemorrhages) were used in this study. Accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative-predictive value for hemorrhage detection were 0.975, 0.983, 0.971, 0.975, 0.793, and 0.997 on training cohort cross-validation and 0.970, 0.981, 0.951, 0.973, 0.829, and 0.993 for the prospective test set. Dice scores for intraparenchymal hemorrhage, epidural/subdural hemorrhage, and SAH were 0.931, 0.863, and 0.772, respectively. A customized deep learning tool is accurate in the detection and quantification of hemorrhage on NCCT. Demonstrated high performance on prospective NCCTs ordered from the emergency department suggests the clinical viability of the proposed deep learning tool.

A scheme for 3-dimensional morphological reconstruction and force inference in the early C. elegans embryo.

In this study, we present a scheme for the reconstruction of cellular morphology and the inference of mechanical forces in the early C. elegans embryo. We have developed and bench-marked a morphological reconstruction scheme that transforms flourescence-based in vivo images of membranes into a point cloud of smoothed surface patches, which facilitates an accurate estimation of membrane curvatures and the angles between membranes. Assuming an isotropic and homogeneous distribution of tensions along individual membranes, we infer a pattern of forces that are 7% deviated from force balance at edges, and 10% deviated from the Young-Laplace relation across membranes. We demonstrate the stability of our inference scheme via a sensitivity analysis, and the reproducibility of our image-analysis and force inference pipelines.

Modeling SNP array ascertainment with Approximate Bayesian Computation for demographic inference

Single nucleotide polymorphisms (SNPs) in commercial arrays have often been discovered in a small number of samples from selected populations. This ascertainment skews patterns of nucleotide diversity and affects population genetic inferences. We propose a demographic inference pipeline that explicitly models the SNP discovery protocol in an Approximate Bayesian Computation (ABC) framework. We simulated genomic regions according to a demographic model incorporating parameters for the divergence of three well-characterized HapMap populations and recreated the SNP distribution of a commercial array by varying the number of haploid samples and the allele frequency cut-off in the given regions. We then calculated summary statistics obtained from both the ascertained and genomic data and inferred ascertainment and demographic parameters. We implemented our pipeline to study the admixture process that gave rise to the present-day Mexican population. Our estimate of the time of admixture is closer to the historical dates than those in previous works which did not consider ascertainment bias. Although the use of whole genome sequences for demographic inference is becoming the norm, there are still underrepresented areas of the world from where only SNP array data are available. Our inference framework is applicable to those cases and will help with the demographic inference.

Dark Energy Survey Year 1 results: the impact of galaxy neighbours on weak lensing cosmology with im3shape

We use a suite of simulated images based on Year 1 of the Dark Energy Survey to explore the impact of galaxy neighbours on shape measurement and shear cosmology. The hoopoe image simulations include realistic blending, galaxy positions, and spatial variations in depth and PSF properties. Using the im3shape maximum-likelihood shape measurement code, we identify four mechanisms by which neighbours can have a non-negligible influence on shear estimation. These effects, if ignored, would contribute a net multiplicative bias of $m \sim 0.03 - 0.09$ in the DES Y1 im3shape catalogue, though the precise impact will be dependent on both the measurement code and the selection cuts applied. This can be reduced to percentage level or less by removing objects with close neighbours, at a cost to the effective number density of galaxies $n_\mathrm{eff}$ of 30%. We use the cosmological inference pipeline of DES Y1 to explore the cosmological implications of neighbour bias and show that omitting blending from the calibration simulation for DES Y1 would bias the inferred clustering amplitude $S_8\equiv \sigma_8 (\Omega _\mathrm{m} /0.3)^{0.5}$ by $2 \sigma$ towards low values. Finally, we use the hoopoe simulations to test the effect of neighbour-induced spatial correlations in the multiplicative bias. We find the impact on the recovered $S_8$ of ignoring such correlations to be subdominant to statistical error at the current level of precision.

Phylogenic inference using alignment-free methods for applications in microbial community surveys using 16s rRNA gene.

The diversity of microbiota is best explored by understanding the phylogenetic structure of the microbial communities. Traditionally, sequence alignment has been used for phylogenetic inference. However, alignment-based approaches come with significant challenges and limitations when massive amounts of data are analyzed. In the recent decade, alignment-free approaches have enabled genome-scale phylogenetic inference. Here we evaluate three alignment-free methods: ACS, CVTree, and Kr for phylogenetic inference with 16s rRNA gene data. We use a taxonomic gold standard to compare the accuracy of alignment-free phylogenetic inference with that of common microbiome-wide phylogenetic inference pipelines based on PyNAST and MUSCLE alignments with FastTree and RAxML. We re-simulate fecal communities from Human Microbiome Project data to evaluate the performance of the methods on datasets with properties of real data. Our comparisons show that alignment-free methods are not inferior to alignment-based methods in giving accurate and robust phylogenic trees. Moreover, consensus ensembles of alignment-free phylogenies are superior to those built from alignment-based methods in their ability to highlight community differences in low power settings. In addition, the overall running times of alignment-based and alignment-free phylogenetic inference are comparable. Taken together our empirical results suggest that alignment-free methods provide a viable approach for microbiome-wide phylogenetic inference.

Genome-Guided Phylo-Transcriptomic Methods and the Nuclear Phylogenetic Tree of the Paniceae Grasses

The past few years have witnessed a paradigm shift in molecular systematics from phylogenetic methods (using one or a few genes) to those that can be described as phylogenomics (phylogenetic inference with entire genomes). One approach that has recently emerged is phylo-transcriptomics (transcriptome-based phylogenetic inference). As in any phylogenetics experiment, accurate orthology inference is critical to phylo-transcriptomics. To date, most analyses have inferred orthology based either on pure sequence similarity or using gene-tree approaches. The use of conserved genome synteny in orthology detection has been relatively under-employed in phylogenetics, mainly due to the cost of sequencing genomes. While current trends focus on the quantity of genes included in an analysis, the use of synteny is likely to improve the quality of ortholog inference. In this study, we combine de novo transcriptome data and sequenced genomes from an economically important group of grass species, the tribe Paniceae, to make phylogenomic inferences. This method, which we call “genome-guided phylo-transcriptomics”, is compared to other recently published orthology inference pipelines, and benchmarked using a set of sequenced genomes from across the grasses. These comparisons provide a framework for future researchers to evaluate the costs and benefits of adding sequenced genomes to transcriptome data sets.

OntoPilot: New Software to Simplify and Accelerate Ontology Development and Deployment

Ontologies (formal descriptions of the concepts and relationships that can exist in knowledge domains) provide the conceptual foundations for modern knowledge representation systems and knowledge bases, and they are a core component of automated reasoning pipelines that infer new facts from existing data. Although a variety of software tools are available for authoring and manipulating ontologies, ontology development remains a technically demanding, time consuming, and often tedious task. In this talk, I will introduce OntoPilot, new software that endeavors to at least partially replace the current patchwork of disparate ontology development tools with a single, unified development framework that is easy to learn and requires minimal knowledge of the technical details of ontology description languages. OntoPilot uses familiar tabular document formats (e.g., spreadsheets) as ontology "source code", which means that domain experts without detailed knowledge of ontology-related technologies can still directly participate in the development process. OntoPilot supports a complete ontology development workflow, including reusing terms from external ontologies, defining new terms, managing releases, and generating end-user documentation. Furthermore, OntoPilot makes it easy to build inference pipelines that use ontologies to generate novel insights from existing data. OntoPilot is open-source software, supports all major operating systems, and is freely available at https://github.com/stuckyb/ontopilot/.

Network-based analysis of transcriptional profiles from chemical perturbations experiments

BackgroundMethods for inference and comparison of biological networks are emerging as powerful tools for the identification of groups of tightly connected genes whose activity may be altered during disease progression or due to chemical perturbations. Connectivity-based comparisons help identify aggregate changes that would be difficult to detect with differential analysis methods comparing individual genes.MethodsIn this study, we describe a pipeline for network comparison and its application to the analysis of gene expression datasets from chemical perturbation experiments, with the goal of elucidating the modes of actions of the profiled perturbations. We apply our pipeline to the analysis of the DrugMatrix and the TG-GATEs, two of the largest toxicogenomics resources available, containing gene expression measurements for model organisms exposed to hundreds of chemical compounds with varying carcinogenicity and genotoxicity.ResultsStarting from chemical-specific transcriptional networks inferred from these data, we show that the proposed comparative analysis of their associated networks identifies groups of chemicals with similar functions and similar carcinogenicity/genotoxicity profiles. We also show that the in-silico annotation by pathway enrichment analysis of the gene modules with a significant gain or loss of connectivity for specific groups of compounds can reveal molecular pathways significantly associated with the chemical perturbations and their likely modes of action.ConclusionsThe proposed pipeline for transcriptional network inference and comparison is highly reproducible and allows grouping chemicals with similar functions and carcinogenicity/genotoxicity profiles. In the context of drug discovery or drug repositioning, the methods presented here could help assign new functions to novel or existing drugs, based on the similarity of their associated network with those built for other known compounds. Additionally, the method has broad applicability beyond the uses here described and could be used as an alternative or as a complement to standard approaches of differential gene expression analysis.

SYGNALing a Red Light for Glioblastoma

A new multiomic network inference pipeline, SYGNAL, integrates patient data with mechanistically accurate transcriptional regulatory networks to predict drug combinations with synergistic anti-proliferative effects on glioblastoma multiforme.

Algorithmic methods to infer the evolutionary trajectories in cancer progression

The genomic evolution inherent to cancer relates directly to a renewed focus on the voluminous next-generation sequencing data and machine learning for the inference of explanatory models of how the (epi)genomic events are choreographed in cancer initiation and development. However, despite the increasing availability of multiple additional -omics data, this quest has been frustrated by various theoretical and technical hurdles, mostly stemming from the dramatic heterogeneity of the disease. In this paper, we build on our recent work on the "selective advantage" relation among driver mutations in cancer progression and investigate its applicability to the modeling problem at the population level. Here, we introduce PiCnIc (Pipeline for Cancer Inference), a versatile, modular, and customizable pipeline to extract ensemble-level progression models from cross-sectional sequenced cancer genomes. The pipeline has many translational implications because it combines state-of-the-art techniques for sample stratification, driver selection, identification of fitness-equivalent exclusive alterations, and progression model inference. We demonstrate PiCnIc's ability to reproduce much of the current knowledge on colorectal cancer progression as well as to suggest novel experimentally verifiable hypotheses.

RockIt: Exploiting Parallelism and Symmetry for MAP Inference in Statistical Relational Models

RockIt is a maximum a-posteriori (MAP) query engine for statistical relational models. MAP inference in graphical models is an optimization problem which can be compiled to integer linear programs (ILPs).We describe several advances in translating MAP queries to ILP instances and present the novel meta-algorithm cutting plane aggregation (CPA). CPA exploits local context-specific symmetries and bundles up sets of linear constraints. The resulting counting constraints lead to more compact ILPs and make the symmetry of the ground model more explicit to state-of-the-art ILP solvers. Moreover, RockIt parallelizes most parts of the MAP inference pipeline taking advantage of ubiquitous shared-memory multi-core architectures. We report on extensive experiments with Markov logic network (MLN) benchmarks showing that RockIt outperforms the state-of-the-art systems Alchemy, Markov TheBeast, and Tuffy both in terms of efficiency and quality of results.

MetaPathways: a modular pipeline for constructing pathway/genome databases from environmental sequence information

BackgroundA central challenge to understanding the ecological and biogeochemical roles of microorganisms in natural and human engineered ecosystems is the reconstruction of metabolic interaction networks from environmental sequence information. The dominant paradigm in metabolic reconstruction is to assign functional annotations using BLAST. Functional annotations are then projected onto symbolic representations of metabolism in the form of KEGG pathways or SEED subsystems.ResultsHere we present MetaPathways, an open source pipeline for pathway inference that uses the PathoLogic algorithm to map functional annotations onto the MetaCyc collection of reactions and pathways, and construct environmental Pathway/Genome Databases (ePGDBs) compatible with the editing and navigation features of Pathway Tools. The pipeline accepts assembled or unassembled nucleotide sequences, performs quality assessment and control, predicts and annotates noncoding genes and open reading frames, and produces inputs to PathoLogic. In addition to constructing ePGDBs, MetaPathways uses MLTreeMap to build phylogenetic trees for selected taxonomic anchor and functional gene markers, converts General Feature Format (GFF) files into concatenated GenBank files for ePGDB construction based on third-party annotations, and generates useful file formats including Sequin files for direct GenBank submission and gene feature tables summarizing annotations, MLTreeMap trees, and ePGDB pathway coverage summaries for statistical comparisons.ConclusionsMetaPathways provides users with a modular annotation and analysis pipeline for predicting metabolic interaction networks from environmental sequence information using an alternative to KEGG pathways and SEED subsystems mapping. It is extensible to genomic and transcriptomic datasets from a wide range of sequencing platforms, and generates useful data products for microbial community structure and function analysis. The MetaPathways software package, installation instructions, and example data can be obtained from http://hallam.microbiology.ubc.ca/MetaPathways.

Abstract 5133: Inference of tumor phylogenetic markers from large copy number datasets.

Abstract Understanding tumors as evolutionary systems is an important area of study with far-reaching implications in diagnostic and treatment paradigms. Computational phylogenetics is a valuable method for inferring tumor evolution in terms of evolutionary trees, phylogenies, where paths in a tree correspond to possible tumor progression pathways. The location of specific cell-types and patient samples in the tree provide information on tumor sub-types and development of heterogeneity. We previously developed a tumor phylogeny inference pipeline for array comparative genome hybridization (aCGH)-based tumor copy number profiles. Steps in the pipeline included extraction of robust progression markers from the data, which could differentiate stages of tumor evolution or the different paths in the tree, and assigning amplification states to the inferred markers in those stages. We introduced a novel multi-sample model for amplicon identification and calling, HMMCNA, which jointly extracted markers from and assigned amplification states to small sets of tumor aCGH profiles. HMMCNA employs a Hidden Markov Model (HMM), a probabilistic model, to classify data into normal and amplified states based on an underlying distribution for the two copy number states and a hidden state space of possible amplification states. We assumed two possible amplification states per sample: normal (0) or amplified (1). Joint segmentation and calling is performed by identifying a most likely sequence of amplification states across all genomic sites probes and samples. This approach limits in the number of samples the HMM can handle since the number of possible hidden amplification states increases exponentially with the number of samples. Here, we present an extension of the approach to handle large datasets. We incorporate a heuristic prior to the HMM classification to reduce the hidden state space by first screening out amplification states not strongly supported at any individual genome coordinates. The introduction of this heuristic reduces the state space on average by 99%. We further reduce the set of possible amplification states based on the frequency of occurrence of the states by only allowing those states occuring at multiple aCGH probes or array genome coordinate. This step accounts for the presence of random noise in the data and gives a further reduction of 80%. We demonstrate the method on a breast tumor aCGH dataset comprising copy number profiles derived from sectioned biopsy samples (NCBI GEO GSE16672, Navin et al., 2010). Our method was able to quickly segment the data into sets of robust normal and amplified segments suitable for downstream phylogeny building. The amplicons inferred carried several known markers of tumor progression. Further steps include tuning the parameters of the HMM to handle noise-levels across different datasets. Citation Format: Ayshwarya Subramanian, Stanley Shackney, Russell Schwartz. Inference of tumor phylogenetic markers from large copy number datasets. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 5133. doi:10.1158/1538-7445.AM2013-5133

AIGO: Towards a unified framework for the Analysis and the Inter-comparison of GO functional annotations

BackgroundIn response to the rapid growth of available genome sequences, efforts have been made to develop automatic inference methods to functionally characterize them. Pipelines that infer functional annotation are now routinely used to produce new annotations at a genome scale and for a broad variety of species. These pipelines differ widely in their inference algorithms, confidence thresholds and data sources for reasoning. This heterogeneity makes a comparison of the relative merits of each approach extremely complex. The evaluation of the quality of the resultant annotations is also challenging given there is often no existing gold-standard against which to evaluate precision and recall.ResultsIn this paper, we present a pragmatic approach to the study of functional annotations. An ensemble of 12 metrics, describing various aspects of functional annotations, is defined and implemented in a unified framework, which facilitates their systematic analysis and inter-comparison. The use of this framework is demonstrated on three illustrative examples: analysing the outputs of state-of-the-art inference pipelines, comparing electronic versus manual annotation methods, and monitoring the evolution of publicly available functional annotations. The framework is part of the AIGO library (http://code.google.com/p/aigo) for the Analysis and the Inter-comparison of the products of Gene Ontology (GO) annotation pipelines. The AIGO library also provides functionalities to easily load, analyse, manipulate and compare functional annotations and also to plot and export the results of the analysis in various formats.ConclusionsThis work is a step toward developing a unified framework for the systematic study of GO functional annotations. This framework has been designed so that new metrics on GO functional annotations can be added in a very straightforward way.

Bayesian inference of ancient human demography from individual genome sequences

Besides their value for biomedicine, individual genome sequences are a rich source of information about human evolution. Here we describe an effort to estimate key evolutionary parameters from sequences for six individuals from diverse human populations. We use a Bayesian, coalescent-based approach to extract information about ancestral population sizes, divergence times, and migration rates from inferred genealogies at many neutrally evolving loci from across the genome. We introduce new methods for accommodating gene flow between populations and integrating over possible phasings of diploid genotypes. We also describe a custom pipeline for genotype inference to mitigate biases from heterogeneous sequencing technologies and coverage levels. Our analysis indicates that the San of Southern Africa diverged from other human populations 108–157 thousand years ago (kya), that Eurasians diverged from an ancestral African population 38–64 kya, and that the effective population size of the ancestors of all modern humans was ~9,000.

DREAM4: Combining Genetic and Dynamic Information to Identify Biological Networks and Dynamical Models

BackgroundCurrent technologies have lead to the availability of multiple genomic data types in sufficient quantity and quality to serve as a basis for automatic global network inference. Accordingly, there are currently a large variety of network inference methods that learn regulatory networks to varying degrees of detail. These methods have different strengths and weaknesses and thus can be complementary. However, combining different methods in a mutually reinforcing manner remains a challenge.MethodologyWe investigate how three scalable methods can be combined into a useful network inference pipeline. The first is a novel t-test–based method that relies on a comprehensive steady-state knock-out dataset to rank regulatory interactions. The remaining two are previously published mutual information and ordinary differential equation based methods (tlCLR and Inferelator 1.0, respectively) that use both time-series and steady-state data to rank regulatory interactions; the latter has the added advantage of also inferring dynamic models of gene regulation which can be used to predict the system's response to new perturbations.Conclusion/SignificanceOur t-test based method proved powerful at ranking regulatory interactions, tying for first out of methods in the DREAM4 100-gene in-silico network inference challenge. We demonstrate complementarity between this method and the two methods that take advantage of time-series data by combining the three into a pipeline whose ability to rank regulatory interactions is markedly improved compared to either method alone. Moreover, the pipeline is able to accurately predict the response of the system to new conditions (in this case new double knock-out genetic perturbations). Our evaluation of the performance of multiple methods for network inference suggests avenues for future methods development and provides simple considerations for genomic experimental design. Our code is publicly available at http://err.bio.nyu.edu/inferelator/.