Multi-omic integration by machine learning (MIMaL)

Saccharomyces cerevisiae Dmo2p is required for the stability and maturation of newly translated Cox2p

Based on available platforms detailing the Saccharomyces cerevisiae mitochondrial proteome and other high-throughput studies, we identified the yeast gene DMO2 as having a profile of genetic and physical interactions that indicate a putative role in mitochondrial respiration. Dmo2p is a homologue to human distal membrane-arm assembly complex protein 1 (DMAC1); both proteins have two conserved cysteines in a Cx2C motif. Here, we localised Dmo2p in the mitochondrial inner membrane with the conserved cysteines facing the intermembrane space. The respiratory deficiency of dmo2 mutants at 37°C led to a reduction in cytochrome c oxidase (COX) activity (COX) and in the formation of cytochrome bc1 complex-COX supercomplexes; dmo2 also has a rapid turnover of Cox2p, the second subunit of the COX complex that harbours the binuclear CuA centre. Moreover, Dmo2p co-immunoprecipitates with Cox2p and components required for maturation of the CuA centre, such as Sco1p and Sco2p. Finally, DMO2 overexpression can suppress cox23 respiratory deficiency, a mutant that has impaired mitochondrial copper homeostasis. Mass spectrometry data unveiled the interaction of Dmo2p with different large molecular complexes, including bc1-COX supercomplexes, the TIM23 machinery and the ADP/ATP nucleotide translocator. Overall, our data suggest that Dmo2p is required for Cox2p maturation, potentially by aiding proteins involved in copper transport and incorporation into Cox2p.

Benchmarking ensemble machine learning algorithms for multi-class, multi-omics data integration in clinical outcome prediction

The complementary information found in different modalities of patient data can aid in more accurate modelling of a patient's disease state and a better understanding of the underlying biological processes of a disease. However, the analysis of multi-modal, multi-omics data presents many challenges. In this work, we compare the performance of a variety of ensemble machine learning (ML) algorithms that are capable of late integration of multi-class data from different modalities. The ensemble methods and their variations tested were (i) a voting ensemble, with hard and soft vote, (ii) a meta learner, and (iii) a multi-modal AdaBoost model using hard vote, soft vote, and meta learner to integrate the modalities on each boosting round, the PB-MVBoost model and a novel application of a mixture of expert's model. These were compared to simple concatenation. We examine these methods using data from an in-house study on hepatocellular carcinoma, plus validation datasets on studies from breast cancer and irritable bowel disease. We develop models that achieve an area under the receiver operating curve of up to 0.85 and find that two boosted methods, PB-MVBoost and AdaBoost with soft vote were the best performing models. We also examine the stability of features selected and the size of the clinical signature. Our work shows that integrating complementary omics and data modalities with effective ensemble ML models enhances accuracy in multi-class clinical outcome predictions and produces more stable predictive features than individual modalities or simple concatenation. We provide recommendations for the integration of multi-modal multi-class data.

Dbi1 is an oxidoreductase and an assembly chaperone for mitochondrial inner membrane proteins

Import and assembly of mitochondrial proteins into multimeric complexes are essential for cellular function. Yet, many steps of these processes and the proteins involved remain unknown. Here, we identify a novel pathway for disulfide bond formation and assembly of mitochondrial inner membrane (IM) proteins. Dbi1, a previously uncharacterized IM protein, interacts with an unassembled pool of Tim17, the central subunit of the presequence translocase of the IM, and is upregulated in cells with increased levels of unassembled Tim17. In the absence of Dbi1, the conformation of the presequence translocase is affected and stability of Tim17 is reduced. Furthermore, Dbi1, through its conserved CxxC motif, is involved in the formation of the disulfide bond in Tim17 in a manner independent of the disulfide relay system, the major oxidation-driven protein import pathway into mitochondria. The substrate spectrum of Dbi1 is not limited to Tim17 but includes at least two more IM proteins, Tim22 and Cox20. We conclude that Dbi1 is a novel oxidoreductase in mitochondria which introduces disulfide bonds into IM proteins and chaperones their assembly into multimeric protein complexes.

Feed-forward loop improves the transient dynamics of an antithetic biological controller

Integral controller is widely used in industry for its capability of endowing perfect adaptation to disturbances. To harness such capability for precise gene expression regulation, synthetic biologists have endeavoured in building biomolecular (quasi-)integral controllers, such as the antithetic integral controller. Despite demonstrated successes, challenges remain with designing the controller for improved transient dynamics and adaptation. Here, we explore and investigate the design principles of alternative RNA-based biological controllers, by modifying an antithetic integral controller with prevalently found natural feed-forward loops (FFL), to improve its transient dynamics and adaptation performance. With model-based analysis, we demonstrate that while the base antithetic controller shows excellent responsiveness and adaptation to system disturbances, incorporating the type-1 incoherent FFL into the base antithetic controller could attenuate the transient dynamics caused by changes in the stimuli, especially in mitigating the undesired overshoot in the output gene expression. Further analysis on the kinetic parameters reveals similar findings to previous studies that the degradation and transcription rates of the circuit RNA species would dominate in shaping the performance of the controllers.

Multiomics Evaluation of Human iPSCs and iPSC-Derived Neurons

Human induced pluripotent stem cells (iPSCs) can be differentiated into neurons, providing living human neurons to model brain diseases. However, it is unclear how different types of molecules work together to regulate stem cell and neuron biology in healthy and disease states. In this study, we conducted integrated proteomics, lipidomics, and metabolomics analyses with confident identification, accurate quantification, and reproducible measurements to compare the molecular profiles of human iPSCs and iPSC-derived neurons. Proteins, lipids, and metabolites related to mitosis, DNA replication, pluripotency, glycosphingolipids, and energy metabolism were highly enriched in iPSCs, whereas synaptic proteins, neurotransmitters, polyunsaturated fatty acids, cardiolipins, and axon guidance pathways were highly enriched in neurons. Mutations in the GRN gene lead to the deficiency of the progranulin (PGRN) protein, which has been associated with various neurodegenerative diseases. Using this multiomics platform, we evaluated the impact of PGRN deficiency on iPSCs and neurons at the whole-cell level. Proteomics, lipidomics, and metabolomics analyses implicated PGRN's roles in neuroinflammation, purine metabolism, and neurite outgrowth, revealing commonly altered pathways related to neuron projection, synaptic dysfunction, and brain metabolism. Multiomics data sets also pointed toward the same hypothesis that neurons seem to be more susceptible to PGRN loss compared to iPSCs, consistent with the neurological symptoms and cognitive impairment from patients carrying inherited GRN mutations.

Multi-omics integration for both single-cell and spatially resolved data based on dual-path graph attention auto-encoder

Single-cell multi-omics integration enables joint analysis at the single-cell level of resolution to provide more accurate understanding of complex biological systems, while spatial multi-omics integration is benefit to the exploration of cell spatial heterogeneity to facilitate more comprehensive downstream analyses. Existing methods are mainly designed for single-cell multi-omics data with little consideration of spatial information and still have room for performance improvement. A reliable multi-omics integration method designed for both single-cell and spatially resolved data is necessary and significant. We propose a multi-omics integration method based on dual-path graph attention auto-encoder (SSGATE). It can construct the neighborhood graphs based on single-cell expression profiles or spatial coordinates, enabling it to process single-cell data and utilize spatial information from spatially resolved data. It can also perform self-supervised learning for integration through the graph attention auto-encoders from two paths. SSGATE is applied to integration of transcriptomics and proteomics, including single-cell and spatially resolved data of various tissues from different sequencing technologies. SSGATE shows better performance and stronger robustness than competitive methods and facilitates downstream analysis.

Multi-dataset Integration and Residual Connections Improve Proteome Prediction from Transcriptomes using Deep Learning

Proteomes are well known to poorly correlate with transcriptomes measured from the same sample. While connected, the complex processes that impact the relationships between transcript and protein quantities remains an open research topic. Many studies have attempted to predict proteomes from transcriptomes with limited success. Here we use publicly available data from the Clinical Proteomics Tumor Analysis Consortium to show that deep learning models designed by neural architecture search (NAS) achieve improved prediction accuracy of proteome quantities from transcriptomics. We find that this benefit is largely due to including a residual connection in the architecture that allows input information to be remembered near the end of the network. Finally, we explore which groups of transcripts are functionally important for protein prediction using model interpretation with SHAP.

SingPro: a knowledge base providing single-cell proteomic data

Single-cell proteomics (SCP) has emerged as a powerful tool for detecting cellular heterogeneity, offering unprecedented insights into biological mechanisms that are masked in bulk cell populations. With the rapid advancements in AI-based time trajectory analysis and cell subpopulation identification, there exists a pressing need for a database that not only provides SCP raw data but also explicitly describes experimental details and protein expression profiles. However, no such database has been available yet. In this study, a database, entitled 'SingPro', specializing in single-cell proteomics was thus developed. It was unique in (a) systematically providing the SCP raw data for both mass spectrometry-based and flow cytometry-based studies and (b) explicitly describing experimental detail for SCP study and expression profile of any studied protein. Anticipating a robust interest from the research community, this database is poised to become an invaluable repository for OMICs-based biomedical studies. Access to SingPro is unrestricted and does not mandate a login at: http://idrblab.org/singpro/.

Simultaneous Multi-Omics Analysis by Direct Infusion Mass Spectrometry (SMAD-MS)

Combined multi-omics analysis of proteomics, polar metabolomics, and lipidomics requires separate liquid chromatography-mass spectrometry (LC-MS) platforms for each omics layer. This requirement for different platforms limits throughput and increases costs, preventing the application of mass spectrometry-based multi-omics to large scale drug discovery or clinical cohorts. Here, we present an innovative strategy for simultaneous multi-omics analysis by direct infusion (SMAD) using one single injection without liquid chromatography. SMAD allows quantification of over 9,000 metabolite m/z features and over 1,300 proteins from the same sample in less than five minutes. We validated the efficiency and reliability of this method and then present two practical applications: mouse macrophage M1/M2 polarization and high throughput drug screening in human 293T cells. Finally, we demonstrate relationships between proteomic and metabolomic data are discovered by machine learning.

From Spatial-Temporal Multiscale Modeling to Application: Bridging the Valley of Death in Industrial Biotechnology

The Valley of Death confronts industrial biotechnology with a significant challenge to the commercialization of products. Fortunately, with the integration of computation, automation and artificial intelligence (AI) technology, the industrial biotechnology accelerates to cross the Valley of Death. The Fourth Industrial Revolution (Industry 4.0) has spurred advanced development of intelligent biomanufacturing, which has evolved the industrial structures in line with the worldwide trend. To achieve this, intelligent biomanufacturing can be structured into three main parts that comprise digitalization, modeling and intellectualization, with modeling forming a crucial link between the other two components. This paper provides an overview of mechanistic models, data-driven models and their applications in bioprocess development. We provide a detailed elaboration of the hybrid model and its applications in bioprocess engineering, including strain design, process control and optimization, as well as bioreactor scale-up. Finally, the challenges and opportunities of biomanufacturing towards Industry 4.0 are also discussed.