MassDash: A Web-Based Dashboard for Data-Independent Acquisition Mass Spectrometry Visualization.

With the increased usage and diversity of methods and instruments being applied to analyze Data-Independent Acquisition (DIA) data, visualization is becoming increasingly important to validate automated software results. Here we present MassDash, a cross-platform DIA mass spectrometry visualization and validation software for comparing features and results across popular tools. MassDash provides a web-based interface and Python package for interactive feature visualizations and summary report plots across multiple automated DIA feature detection tools, including OpenSwath, DIA-NN, and dreamDIA. Furthermore, MassDash processes peptides on the fly, enabling interactive visualization of peptides across dozens of runs simultaneously on a personal computer. MassDash supports various multidimensional visualizations across retention time, ion mobility, m/z, and intensity, providing additional insights into the data. The modular framework is easily extendable, enabling rapid algorithm development of novel peak-picker techniques, such as deep-learning-based approaches and refinement of existing tools. MassDash is open-source under a BSD 3-Clause license and freely available at https://github.com/Roestlab/massdash, and a demo version can be accessed at https://massdash.streamlit.app.

Longitudinal cytokine and multi-modal health data of an extremely severe ME/CFS patient with HSD reveals insights into immunopathology, and disease severity.

Introduction: Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) presents substantial challenges in patient care due to its intricate multisystem nature, comorbidities, and global prevalence. The heterogeneity among patient populations, coupled with the absence of FDA-approved diagnostics and therapeutics, further complicates research into disease etiology and patient managment. Integrating longitudinal multi-omics data with clinical, health,textual, pharmaceutical, and nutraceutical data offers a promising avenue to address these complexities, aiding in the identification of underlying causes and providing insights into effective therapeutics and diagnostic strategies. Methods: This study focused on an exceptionally severe ME/CFS patient with hypermobility spectrum disorder (HSD) during a period of marginal symptom improvements. Longitudinal cytokine profiling was conducted alongside the collection of extensive multi-modal health data to explore the dynamic nature of symptoms, severity, triggers, and modifying factors. Additionally, an updated severity assessment platform and two applications, ME-CFSTrackerApp and LexiTime, were introduced to facilitate real-time symptom tracking and enhance patient-physician/researcher communication, and evaluate response to medical intervention. Results: Longitudinal cytokine profiling revealed the significance of Th2-type cytokines and highlighted synergistic activities between mast cells and eosinophils, skewing Th1 toward Th2 immune responses in ME/CFS pathogenesis, particularly in cognitive impairment and sensorial intolerance. This suggests a potentially shared underlying mechanism with major ME/CFS comorbidities such as HSD, Mast cell activation syndrome, postural orthostatic tachycardia syndrome (POTS), and small fiber neuropathy. Additionally, the data identified potential roles of BCL6 and TP53 pathways in ME/CFS etiology and emphasized the importance of investigating adverse reactions to medication and supplements and drug interactions in ME/CFS severity and progression. Discussion: Our study advocates for the integration of longitudinal multi-omics with multi-modal health data and artificial intelligence (AI) techniques to better understand ME/CFS and its major comorbidities. These findings highlight the significance of dysregulated Th2-type cytokines in patient stratification and precision medicine strategies. Additionally, our results suggest exploring the use of low-dose drugs with partial agonist activity as a potential avenue for ME/CFS treatment. This comprehensive approach emphasizes the importance of adopting a patient-centered care approach to improve ME/CFS healthcare management, disease severity assessment, and personalized medicine. Overall, these findings contribute to our understanding of ME/CFS and offer avenues for future research and clinical practice.

Cytidine deaminases APOBEC3C and APOBEC3D promote DNA replication stress resistance in pancreatic cancer cells.

Gemcitabine is a potent inhibitor of DNA replication and is a mainstay therapeutic for diverse cancers, particularly pancreatic ductal adenocarcinoma (PDAC). However, most tumors remain refractory to gemcitabine therapies. Here, to define the cancer cell response to gemcitabine, we performed genome-scale CRISPR-Cas9 chemical-genetic screens in PDAC cells and found selective loss of cell fitness upon disruption of the cytidine deaminases APOBEC3C and APOBEC3D. Following gemcitabine treatment, APOBEC3C and APOBEC3D promote DNA replication stress resistance and cell survival by deaminating cytidines in the nuclear genome to ensure DNA replication fork restart and repair in PDAC cells. We provide evidence that the chemical-genetic interaction between APOBEC3C or APOBEC3D and gemcitabine is absent in nontransformed cells but is recapitulated across different PDAC cell lines, in PDAC organoids and in PDAC xenografts. Thus, we uncover roles for APOBEC3C and APOBEC3D in DNA replication stress resistance and offer plausible targets for improving gemcitabine-based therapies for PDAC.

Computational Expansion of High-Resolution-MSn Spectral Libraries.

Commonly, in MS-based untargeted metabolomics, some metabolites cannot be confidently identified due to ambiguities in resolving isobars and structurally similar species. To address this, analytical techniques beyond traditional MS2 analysis, such as MSn fragmentation, can be applied to probe metabolites for additional structural information. In MSn fragmentation, recursive cycles of activation are applied to fragment ions originating from the same precursor ion detected on an MS1 spectrum. This resonant-type collision-activated dissociation (CAD) can yield information that cannot be ascertained from MS2 spectra alone, which helps improve the performance of metabolite identification workflows. However, most approaches for metabolite identification require mass-to-charge (m/z) values measured with high resolution, as this enables the determination of accurate mass values. Unfortunately, high-resolution-MSn spectra are relatively rare in spectral libraries. Here, we describe a computational approach to generate a database of high-resolution-MSn spectra by converting existing low-resolution-MSn spectra using complementary high-resolution-MS2 spectra generated by beam-type CAD. Using this method, we have generated a database, derived from the NIST20 MS/MS database, of MSn spectral trees representing 9637 compounds and 19386 precursor ions where at least 90% of signal intensity was converted from low-to-high resolution.

Achieving quantitative reproducibility in label-free multisite DIA experiments through multirun alignment.

DIA is a mainstream method for quantitative proteomics, but consistent quantification across multiple LC-MS/MS instruments remains a bottleneck in parallelizing data acquisition. One reason for this inconsistency and missing quantification is the retention time shift which current software does not adequately address for runs from multiple sites. We present multirun chromatogram alignment strategies to map peaks across columns, including the traditional reference-based Star method, and two novel approaches: MST and Progressive alignment. These reference-free strategies produce a quantitatively accurate data-matrix, even from heterogeneous multi-column studies. Progressive alignment also generates merged chromatograms from all runs which has not been previously achieved for LC-MS/MS data. First, we demonstrate the effectiveness of multirun alignment strategies on a gold-standard annotated dataset, resulting in a threefold reduction in quantitation error-rate compared to non-aligned DIA results. Subsequently, on a multi-species dataset that DIAlignR effectively controls the quantitative error rate, improves precision in protein measurements, and exhibits conservative peak alignment. We next show that the MST alignment reduces cross-site CV by 50% for highly abundant proteins when applied to a dataset from 11 different LC-MS/MS setups. Finally, the reanalysis of 949 plasma runs with multirun alignment revealed a more than 50% increase in insulin resistance (IR) and respiratory viral infection (RVI) proteins, identifying 11 and 13 proteins respectively, compared to prior analysis without it. The three strategies are implemented in our DIAlignR workflow (>2.3) and can be combined with linear, non-linear, or hybrid pairwise alignment.

Oxidized Low-Density Lipoprotein Accumulation Suppresses Glycolysis and Attenuates the Macrophage Inflammatory Response by Diverting Transcription from the HIF-1α to the Nrf2 Pathway.

Lipid accumulation in macrophages (Mφs) is a hallmark of atherosclerosis, yet how lipid accumulation affects inflammatory responses through rewiring of Mφ metabolism is poorly understood. We modeled lipid accumulation in cultured wild-type mouse thioglycolate-elicited peritoneal Mφs and bone marrow-derived Mφs with conditional (Lyz2-Cre) or complete genetic deficiency of Vhl, Hif1a, Nos2, and Nfe2l2. Transfection studies employed RAW264.7 cells. Mφs were cultured for 24 h with oxidized low-density lipoprotein (oxLDL) or cholesterol and then were stimulated with LPS. Transcriptomics revealed that oxLDL accumulation in Mφs downregulated inflammatory, hypoxia, and cholesterol metabolism pathways, whereas the antioxidant pathway, fatty acid oxidation, and ABC family proteins were upregulated. Metabolomics and extracellular metabolic flux assays showed that oxLDL accumulation suppressed LPS-induced glycolysis. Intracellular lipid accumulation in Mφs impaired LPS-induced inflammation by reducing both hypoxia-inducible factor 1-α (HIF-1α) stability and transactivation capacity; thus, the phenotype was not rescued in Vhl-/- Mφs. Intracellular lipid accumulation in Mφs also enhanced LPS-induced NF erythroid 2-related factor 2 (Nrf2)-mediated antioxidative defense that destabilizes HIF-1α, and Nrf2-deficient Mφs resisted the inhibitory effects of lipid accumulation on glycolysis and inflammatory gene expression. Furthermore, oxLDL shifted NADPH consumption from HIF-1α- to Nrf2-regulated apoenzymes. Thus, we postulate that repurposing NADPH consumption from HIF-1α to Nrf2 transcriptional pathways is critical in modulating inflammatory responses in Mφs with accumulated intracellular lipid. The relevance of our in vitro models was established by comparative transcriptomic analyses, which revealed that Mφs cultured with oxLDL and stimulated with LPS shared similar inflammatory and metabolic profiles with foamy Mφs derived from the atherosclerotic mouse and human aorta.

Discovery and Structural Characterization of Small Molecule Binders of the Human CTLH E3 Ligase Subunit GID4.

Targeted protein degradation (TPD) strategies exploit bivalent small molecules to bridge substrate proteins to an E3 ubiquitin ligase to induce substrate degradation. Few E3s have been explored as degradation effectors due to a dearth of E3-binding small molecules. We show that genetically induced recruitment to the GID4 subunit of the CTLH E3 complex induces protein degradation. An NMR-based fragment screen followed by structure-guided analog elaboration identified two binders of GID4, 16 and 67, with Kd values of 110 and 17 µM in vitro. A parallel DNA-encoded library (DEL) screen identified five binders of GID4, the best of which, 88, had a Kd of 5.6 µM in vitro and an EC50 of 558 nM in cells with strong selectivity for GID4. X-ray co-structure determination revealed the basis for GID4-small molecule interactions. These results position GID4-CTLH as an E3 for TPD and provide candidate scaffolds for high-affinity moieties that bind GID4.

Improving Phosphoproteomics Profiling Using Data-Independent Mass Spectrometry.

Mass spectrometry-based profiling of the phosphoproteome is a powerful method of identifying phosphorylation events at a systems level. Most phosphoproteomics studies have used data-dependent acquisition (DDA) mass spectrometry as their method of choice. In this Perspective, we review some recent studies benchmarking DDA and DIA methods for phosphoproteomics and discuss data analysis options for DIA phosphoproteomics. In order to evaluate the impact of data-dependent and data-independent acquisition (DIA) on identification and quantification, we analyze a previously published phosphopeptide-enriched data set consisting of 10 replicates acquired by DDA and DIA each. We find that though more unique identifications are made in DDA data, phosphopeptides are identified more consistently across replicates in DIA. We further discuss the challenges of identifying chromatographically coeluting phosphopeptide isomers and investigate the impact on reproducibility of identifying high-confidence site-localized phosphopeptides in replicates.

Proteomic Characterization of a Candidate Polygenic Driver of Metabolism in Non-small Cell Lung Cancer.

Proteome analysis revealed signatures of co-expressed upregulated metabolism proteins highly conserved between primary and non-small cell lung cancer (NSCLC) patient-derived xenograft tumors (Li et al. 2014, Nat. Communications 5:5469). The C10 signature is encoded by seven genes (ADSS, ATP2A2, CTPS1, IMPDH2, PKM2, PTGES3, SGPL1) and DNA alterations in C10-encoding genes are associated with longer survival in a subset of NSCLC. To explore the C10 signature as an oncogenic driver and address potential mechanisms of action, C10 protein expression and protein-protein interactions were determined. In independent NSCLC cohorts, the coordinated expression of C10 proteins was significant and mutations in C10 genes were associated with better outcome. Affinity purification-mass spectrometry and in vivo proximity-based biotin identification defined a C10 interactome involving 667 proteins including candidate drug targets and clusters associated with glycolysis, calcium homeostasis, and nucleotide and sphingolipid metabolism. DNA alterations in genes encoding C10 interactome components were also found to be associated with better survival. These data support the notion that the coordinated upregulation of the C10 signature impinges metabolic processes that collectively function as an oncogenic driver in NSCLC.

DIAMetAlyzer allows automated false-discovery rate-controlled analysis for data-independent acquisition in metabolomics.

The extraction of meaningful biological knowledge from high-throughput mass spectrometry data relies on limiting false discoveries to a manageable amount. For targeted approaches in metabolomics a main challenge is the detection of false positive metabolic features in the low signal-to-noise ranges of data-independent acquisition results and their filtering. Another factor is that the creation of assay libraries for data-independent acquisition analysis and the processing of extracted ion chromatograms have not been automated in metabolomics. Here we present a fully automated open-source workflow for high-throughput metabolomics that combines data-dependent and data-independent acquisition for library generation, analysis, and statistical validation, with rigorous control of the false-discovery rate while matching manual analysis regarding quantification accuracy. Using an experimentally specific data-dependent acquisition library based on reference substances allows for accurate identification of compounds and markers from data-independent acquisition data in low concentrations, facilitating biomarker quantification.

Trapped Ion Mobility Spectrometry Reduces Spectral Complexity in Mass Spectrometry-Based Proteomics.

In bottom-up mass spectrometry-based proteomics, deep proteome coverage is limited by high cofragmentation rates. Cofragmentation occurs when more than one analyte is isolated by the quadrupole and the subsequent fragmentation event produces fragment ions of heterogeneous origin. One strategy to reduce cofragmentation rates is through effective peptide separation techniques such as chromatographic separation and, the more recently popularized, ion mobility (IM) spectrometry, which separates peptides by their collisional cross section. Here, we use a computational model to investigate the capability of the trapped IM spectrometry (TIMS) device at effectively separating peptide ions and quantify the separation power of the TIMS device in the context of a parallel accumulation-serial fragmentation (PASEF) workflow. We found that TIMS separation increases the number of interference-free MS1 peptide features 9.2-fold, while decreasing the average peptide density in precursor spectra 6.5-fold. In a data-dependent acquisition PASEF workflow, IM separation increases the number of spectra without cofragmentation by a factor of 4.1 and the number of high-quality spectra 17-fold. Using a categorical model, we estimate that this observed decrease in spectral complexity results in an increased likelihood for peptide spectral matches, which may improve peptide identification rates. In the context of a data-independent acquisition workflow, the reduction in spectral complexity resulting from IM separation is estimated to be equivalent to a 4-fold decrease in the isolation window width (from 25 to 6.5 Da). Our study demonstrates that TIMS separation decreases spectral complexity by reducing cofragmentation rates, suggesting that TIMS separation may contribute toward the high identification rates observed in PASEF workflows.

Intensive lactation among women with recent gestational diabetes significantly alters the early postpartum circulating lipid profile: the SWIFT study.

BACKGROUND: Women with a history of gestational diabetes mellitus (GDM) have a 7-fold higher risk of developing type 2 diabetes (T2D). It is estimated that 20-50% of women with GDM history will progress to T2D within 10 years after delivery. Intensive lactation could be negatively associated with this risk, but the mechanisms behind a protective effect remain unknown. METHODS: In this study, we utilized a prospective GDM cohort of 1010 women without T2D at 6-9 weeks postpartum (study baseline) and tested for T2D onset up to 8 years post-baseline (n=980). Targeted metabolic profiling was performed on fasting plasma samples collected at both baseline and follow-up (1-2 years post-baseline) during research exams in a subset of 350 women (216 intensive breastfeeding, IBF vs. 134 intensive formula feeding or mixed feeding, IFF/Mixed). The relationship between lactation intensity and circulating metabolites at both baseline and follow-up were evaluated to discover underlying metabolic responses of lactation and to explore the link between these metabolites and T2D risk. RESULTS: We observed that lactation intensity was strongly associated with decreased glycerolipids (TAGs/DAGs) and increased phospholipids/sphingolipids at baseline. This lipid profile suggested decreased lipogenesis caused by a shift away from the glycerolipid metabolism pathway towards the phospholipid/sphingolipid metabolism pathway as a component of the mechanism underlying the benefits of lactation. Longitudinal analysis demonstrated that this favorable lipid profile was transient and diminished at 1-2 years postpartum, coinciding with the cessation of lactation. Importantly, when stratifying these 350 women by future T2D status during the follow-up (171 future T2D vs. 179 no T2D), we discovered that lactation induced robust lipid changes only in women who did not develop incident T2D. Subsequently, we identified a cluster of metabolites that strongly associated with future T2D risk from which we developed a predictive metabolic signature with a discriminating power (AUC) of 0.78, superior to common clinical variables (i.e., fasting glucose, AUC 0.56 or 2-h glucose, AUC 0.62). CONCLUSIONS: In this study, we show that intensive lactation significantly alters the circulating lipid profile at early postpartum and that women who do not respond metabolically to lactation are more likely to develop T2D. We also discovered a 10-analyte metabolic signature capable of predicting future onset of T2D in IBF women. Our findings provide novel insight into how lactation affects maternal metabolism and its link to future diabetes onset. TRIAL REGISTRATION: ClinicalTrials.gov NCT01967030 .

PRMT5 inhibition disrupts splicing and stemness in glioblastoma.

Glioblastoma (GBM) is a deadly cancer in which cancer stem cells (CSCs) sustain tumor growth and contribute to therapeutic resistance. Protein arginine methyltransferase 5 (PRMT5) has recently emerged as a promising target in GBM. Using two orthogonal-acting inhibitors of PRMT5 (GSK591 or LLY-283), we show that pharmacological inhibition of PRMT5 suppresses the growth of a cohort of 46 patient-derived GBM stem cell cultures, with the proneural subtype showing greater sensitivity. We show that PRMT5 inhibition causes widespread disruption of splicing across the transcriptome, particularly affecting cell cycle gene products. We identify a GBM splicing signature that correlates with the degree of response to PRMT5 inhibition. Importantly, we demonstrate that LLY-283 is brain-penetrant and significantly prolongs the survival of mice with orthotopic patient-derived xenografts. Collectively, our findings provide a rationale for the clinical development of brain penetrant PRMT5 inhibitors as treatment for GBM.

diaPASEF: parallel accumulation-serial fragmentation combined with data-independent acquisition.

Data-independent acquisition modes isolate and concurrently fragment populations of different precursors by cycling through segments of a predefined precursor m/z range. Although these selection windows collectively cover the entire m/z range, overall, only a few per cent of all incoming ions are isolated for mass analysis. Here, we make use of the correlation of molecular weight and ion mobility in a trapped ion mobility device (timsTOF Pro) to devise a scan mode that samples up to 100% of the peptide precursor ion current in m/z and mobility windows. We extend an established targeted data extraction workflow by inclusion of the ion mobility dimension for both signal extraction and scoring and thereby increase the specificity for precursor identification. Data acquired from whole proteome digests and mixed organism samples demonstrate deep proteome coverage and a high degree of reproducibility as well as quantitative accuracy, even from 10 ng sample amounts.

Underlying dyslipidemia postpartum in women with a recent GDM pregnancy who develop type 2 diabetes.

Approximately, 35% of women with Gestational Diabetes (GDM) progress to Type 2 Diabetes (T2D) within 10 years. However, links between GDM and T2D are not well understood. We used a well-characterised GDM prospective cohort of 1035 women following up to 8 years postpartum. Lipidomics profiling covering >1000 lipids was performed on fasting plasma samples from participants 6-9 week postpartum (171 incident T2D vs. 179 controls). We discovered 311 lipids positively and 70 lipids negatively associated with T2D risk. The upregulation of glycerolipid metabolism involving triacylglycerol and diacylglycerol biosynthesis suggested activated lipid storage before diabetes onset. In contrast, decreased sphingomyelines, hexosylceramide and lactosylceramide indicated impaired sphingolipid metabolism. Additionally, a lipid signature was identified to effectively predict future diabetes risk. These findings demonstrate an underlying dyslipidemia during the early postpartum in those GDM women who progress to T2D and suggest endogenous lipogenesis may be a driving force for future diabetes onset.

Amino acid and lipid metabolism in post-gestational diabetes and progression to type 2 diabetes: A metabolic profiling study.

BACKGROUND: Women with a history of gestational diabetes mellitus (GDM) have a 7-fold higher risk of developing type 2 diabetes (T2D) during midlife and an elevated risk of developing hypertension and cardiovascular disease. Glucose tolerance reclassification after delivery is recommended, but fewer than 40% of women with GDM are tested. Thus, improved risk stratification methods are needed, as is a deeper understanding of the pathology underlying the transition from GDM to T2D. We hypothesize that metabolites during the early postpartum period accurately distinguish risk of progression from GDM to T2D and that metabolite changes signify underlying pathophysiology for future disease development. METHODS AND FINDINGS: The study utilized fasting plasma samples collected from a well-characterized prospective research study of 1,035 women diagnosed with GDM. The cohort included racially/ethnically diverse pregnant women (aged 20-45 years-33% primiparous, 37% biparous, 30% multiparous) who delivered at Kaiser Permanente Northern California hospitals from 2008 to 2011. Participants attended in-person research visits including 2-hour 75-g oral glucose tolerance tests (OGTTs) at study baseline (6-9 weeks postpartum) and annually thereafter for 2 years, and we retrieved diabetes diagnoses from electronic medical records for 8 years. In a nested case-control study design, we collected fasting plasma samples among women without diabetes at baseline (n = 1,010) to measure metabolites among those who later progressed to incident T2D or did not develop T2D (non-T2D). We studied 173 incident T2D cases and 485 controls (pair-matched on BMI, age, and race/ethnicity) to discover metabolites associated with new onset of T2D. Up to 2 years post-baseline, we analyzed samples from 98 T2D cases with 239 controls to reveal T2D-associated metabolic changes. The longitudinal analysis tracked metabolic changes within individuals from baseline to 2 years of follow-up as the trajectory of T2D progression. By building prediction models, we discovered a distinct metabolic signature in the early postpartum period that predicted future T2D with a median discriminating power area under the receiver operating characteristic curve of 0.883 (95% CI 0.820-0.945, p < 0.001). At baseline, the most striking finding was an overall increase in amino acids (AAs) as well as diacyl-glycerophospholipids and a decrease in sphingolipids and acyl-alkyl-glycerophospholipids among women with incident T2D. Pathway analysis revealed up-regulated AA metabolism, arginine/proline metabolism, and branched-chain AA (BCAA) metabolism at baseline. At follow-up after the onset of T2D, up-regulation of AAs and down-regulation of sphingolipids and acyl-alkyl-glycerophospholipids were sustained or strengthened. Notably, longitudinal analyses revealed only 10 metabolites associated with progression to T2D, implicating AA and phospholipid metabolism. A study limitation is that all of the analyses were performed with the same cohort. It would be ideal to validate our findings in an independent longitudinal cohort of women with GDM who had glucose tolerance tested during the early postpartum period. CONCLUSIONS: In this study, we discovered a metabolic signature predicting the transition from GDM to T2D in the early postpartum period that was superior to clinical parameters (fasting plasma glucose, 2-hour plasma glucose). The findings suggest that metabolic dysregulation, particularly AA dysmetabolism, is present years prior to diabetes onset, and is revealed during the early postpartum period, preceding progression to T2D, among women with GDM. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT01967030.

Publisher Correction: OpenSWATH enables automated, targeted analysis of data-independent acquisition MS data.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Machine Learning in Mass Spectrometric Analysis of DIA Data.

Liquid Chromatography coupled to Tandem Mass Spectrometry (LC-MS/MS) based methods are currently the top choice for high-throughput, quantitative measurements of the proteome. While traditional proteomics LC-MS/MS methods can suffer from issues such as low reproducibility and quantitative accuracy due to its stochastic nature, recent improvements in acquisition protocols have resulted in methods that can overcome these challenges. Data-independent acquisition (DIA) is a novel mass spectrometric method that does so by using a deterministic acquisition strategy. These new approaches will allow researchers to apply MS on more complex samples, however, existing heuristic and expert-knowledge based methods will struggle with keeping pace of the increasing complexity of the resulting data. Deep learning (DL) based methods have been shown to be more adept at handling large amounts of complex data than traditional methods in many other fields, such as computer vision and natural language processing. Proteomics is also entering a phase where the size and complexity of the data will require us to look towards scalable and data-driven DL pipelines.