Framework For Discovery Research Articles

The Semantic Adjacency Criterion in Time Intervals Mining

We propose a new pruning constraint when mining frequent temporal patterns to be used as classification and prediction features, the Semantic Adjacency Criterion [SAC], which filters out temporal patterns that contain potentially semantically contradictory components, exploiting each medical domain’s knowledge. We have defined three SAC versions and tested them within three medical domains (oncology, hepatitis, diabetes) and a frequent-temporal-pattern discovery framework. Previously, we had shown that using SAC enhances the repeatability of discovering the same temporal patterns in similar proportions in different patient groups within the same clinical domain. Here, we focused on SAC’s computational implications for pattern discovery, and for classification and prediction, using the discovered patterns as features, by four different machine-learning methods: Random Forests, Naïve Bayes, SVM, and Logistic Regression. Using SAC resulted in a significant reduction, across all medical domains and classification methods, of up to 97% in the number of discovered temporal patterns, and in the runtime of the discovery process, of up to 98%. Nevertheless, the highly reduced set of only semantically transparent patterns, when used as features, resulted in classification and prediction models whose performance was at least as good as the models resulting from using the complete temporal-pattern set.

A Multimodal Single Cell Atlas of Adult Murine Haematopoiesis Reveals Nine Differentiation Trajectories and Epigenetic Regulatory Programs

Single cell genomics has revolutionised our understanding of differentiating systems and led to a reinterpretation of haematopoiesis, moving from a step-wise hierarchical tree model towards more continous differentiation landscapes. Recent single cell RNA sequencing (scRNA-seq) studies of Lineage- ckit+ (LK) mouse bone marrow progenitors have defined a highly granular landscape of differentiation from multipotent stem cells to committed progenitors in 8 different lineages and have served as a valuable reference landscape for comparisons with perturbation states. While these landscapes have permitted exploration and interrogation of early haematopoietic progenitors, as yet, no map of mouse total bone marrow haematopoiesis exists at single cell resolution, limiting efforts to fully understand the gene programs which specify and the cell surface phenotypes which identify maturing haematopoietic lineages nor the connection between early bifurcation decisions in progenitor cells and mature cellular outputs. To answer these questions, we performed droplet-based scRNA-seq and cell surface proteomics on whole bone marrow from 4 male and 4 female 12-week old C57Bl6 mice, using a panel of 138 antibodies against cell surface antigens. To increase representation of rare progenitor populations, we complemented total bone marrow mononuclear cells with c-kit enrichment and FACS sorted LK and LK sca1-positive (LSK) populations. After quality control filtering the single cell transcriptomes and proteomes, over 198,000 single cells were analysed separately before multimodal integration into a single UMAP embedding. Addition of cell surface proteome information to the scRNA-seq data through multimodal integration produced a representation with greater resolution of haematopoietic cell types, including improved resolution of T-cell subtypes and separation of a rare population of eosinophil progenitors which were previously indistinguishable from basophil progenitors. This integrated landscape of mouse haematopoiesis was then clustered and annotated though an iterative process of label transfer and manual annotation into 41 cell type states (Figure 1). Using cell fate analysis, which utilises transcriptional similarity and pseudotemporal ordering to estimate the probability of a single cell's commitment to defined terminal states, we were able to resolve cell fate probabilities towards nine haematopoietic lineages (erythroid, neutrophil, megakaryocyte, lymphoid, monocyte, dendritic cell (DC), plasmacytoid dendritic cell (pDC), basophil and eosinophil (Figure 2)). From these probabilities we defined nine trajectories from the earliest uncommitted progenitors to terminally differentiated populations, including a pDC trajectory involving both lymphoid and common DC progenitors. Analysis of trajectory-based gene expression trends provides a framework for discovery of lineage drivers as well as shared gene expression programs across multiple lineages. Leveraging cell surface proteome data permits the use of ‘in-silico-FACS’ to describe phenotypes of cell populations which exist at cell fate branch-points, for further in-vitro characterisation. To demonstrate the utility of our atlas as a reference, we projected perturbation datasets from 7 pre-leukaemic mouse models onto the atlas, allowing quantitation of mutation-associated cellular and tissue-scale alterations. Finally, we have leveraged the transcription-start site enrichment of 5' scRNA-seq to compute enhancer RNA (eRNA) expression in our atlas, allowing the identification of dynamic usage of specific enhancer elements during haematopoietic specification and thus providing a novel framework for correlation of eRNA expression (as a proxy for active distal chromatin regions) with gene expression trends along specific haematopoietic lineages.

Collaborative analysis for drug discovery by federated learning on non-IID data.

With the increasing availability of large-scale QSAR (Quantitative Structure-Activity Relationship) datasets, collaborative analysis has become a promising approach for drug discovery. Traditional centralized analysis which typically concentrates data on a central server for training faces challenges such as data privacy and security. Distributed analysis such as federated learning offers a solution by enabling collaborative model training without sharing raw data. However, it may fail when the training data in the local devices are non-independent and identically distributed (non-IID). In this paper, we propose a novel framework for collaborative drug discovery using federated learning on non-IID datasets. We address the difficulty of training on non-IID data by globally sharing a small subset of data among all institutions. Our framework allows multiple institutions to jointly train a robust predictive model while preserving the privacy of their individual data. We leverage the federated learning paradigm to distribute the model training process across local devices, eliminating the need for data exchange. The experimental results on 15 benchmark datasets demonstrate that the proposed method achieves competitive predictive accuracy to centralized analysis while respecting data privacy. Moreover, our framework offers benefits such as reduced data transmission and enhanced scalability, making it suitable for large-scale collaborative drug discovery efforts.

A Privacy-Preserving and Reputation-Based Truth Discovery Framework in Mobile Crowdsensing

In mobile crowdsensing (MCS), truth discovery (TD) plays an important role in sensing task completion. Most of the existing studies focus on the privacy preservation of mobile users, and the reliability of mobile users is evaluated by their weights which are calculated based on the submitted sensing data. However, if mobile users are unreliable, the submitted sensing data and their weights are also unreliable, which may influence the accuracy of the ground truths of sensing tasks. Therefore, this paper proposes a privacy-preserving and reputation-based truth discovery framework named PRTD which can generate the ground truths of sensing tasks with high accuracy while preserving privacy. Specifically, we first preserve sensing data privacy, weight privacy, and reputation value privacy by utilizing the Paillier algorithm and Pedersen commitment. Then, to verify whether the reputation values of mobile users are tampered with and select mobile users that satisfy the corresponding reputation requirements, we design a privacy-preserving reputation verification algorithm based on reputation commitment and zero-knowledge proof and propose a concept of reliability level to select mobile users. Finally, a general TD algorithm with reliability level is presented to improve the accuracy of the ground truths of sensing tasks. Moreover, theoretical analysis and performance evaluation are conducted, and the evaluation results demonstrate that the PRTD framework outperforms the existing TD frameworks in several evaluation metrics in the synthetic dataset and real-world dataset.

Active relation discovery: Towards general and label-aware open relation extraction

Open Relation Extraction (OpenRE) aims to discover novel relations from open domains. Previous OpenRE methods mainly suffer from two problems: (1) Insufficient capacity to discriminate between known and novel relations. When extending conventional test settings to a more general setting where test data might also come from seen classes, existing approaches have a significant performance decline. (2) Secondary labeling must be performed before practical application. Existing methods cannot label human-readable and meaningful types for novel relations, which is urgently required by the downstream tasks. To address these issues, we propose the Active Relation Discovery (ARD) framework, which utilizes relational outlier detection for discriminating known and novel relations and involves active learning for labeling novel relations. Extensive experiments on three real-world datasets show that ARD significantly outperforms previous state-of-the-art methods on both conventional and our proposed general OpenRE settings. The source code and datasets will be available for reproducibility.

Model discovery of compartmental models with Graph-Supported Neural Networks

In this proposal, our objective is to create a neural network for the discovery of models using compartmental models as systems of Ordinary Differential Equations (ODEs), employing Graph-Supported Neural Networks (GSNN). We design the GSNN as a graph of transition functions of the structure of the model to ensure that the three properties of the ODE solution are satisfied: additivity to one, positivity, and boundness. Rather than directly estimating the solution, these neural networks approximate the transition functions. We present theoretical evidence substantiating that our GSNN maintains these properties, along with approximation outcomes from both simulated and real-world data. These outcomes demonstrate that our GSNN outperforms the non-graph-supported approach for these issues, with instances where it even surpasses full model-based solutions and Recurrent Neural Networks. Furthermore, an evolutionary algorithm successfully generated a consistent model using the data, offering a comprehensive framework for neural network-based model discovery in these scenarios.

Temporal Precursor Discovery Using Long Short-Term Memory with Feature Attention

The continuous growth of demand on commercial airlines has made it crucial to guarantee the safety of airspace operations. Although adverse events are rare, once they happen, they can cause unpredictable risky factors and degrade airspace efficiency. Thus, studying historical air traffic data to discover precursors, features, or events that contribute to the occurrence of the adverse event in the future is important and has gained interest in recent years. In this paper, a novel and real-time applicable temporal precursor discovery (TPD) framework based on the long short-term memory neural network and the feature attention mechanism is proposed. The feature attention mechanism enables the framework to pay attention to certain features at a certain time, and the attention score is defined as the temporal precursor. The temporal precursor reflects the rationale behind the neural network’s prediction at each time step, providing a data-driven explanation of how the adverse event occurs. The proposed TPD framework was tested with real air traffic data and weather data recorded at Incheon International Airport in South Korea in 2019.

Discovering interpretable Lagrangian of dynamical systems from data

A complete understanding of physical systems requires models that are accurate and obey natural conservation laws. Recent trends in representation learning involve learning Lagrangian from data rather than the direct discovery of governing equations of motion. The generalization of equation discovery techniques has huge potential; however, existing Lagrangian discovery frameworks are black-box in nature. This raises a concern about the reusability of the discovered Lagrangian. In this article, we propose a novel data-driven machine-learning algorithm to automate the discovery of interpretable Lagrangian from data. The Lagrangian is discovered in interpretable forms, which also allows the automated discovery of conservation laws and governing equations of motion. The architecture of the proposed framework is designed in such a way that it allows learning the Lagrangian from a subset of the underlying domain and then generalizing for an infinite-dimensional system. The fidelity of the proposed framework is exemplified using examples described by systems of ordinary differential equations and partial differential equations where the Lagrangian and conserved quantities are known.

Atosiban and Rutin exhibit anti-mycobacterial activity - An integrated computational and biophysical insight toward drug repurposing strategy against Mycobacterium tuberculosis targeting its essential enzyme HemD

With the advancements of high throughput computational screening procedures, drug repurposing became the privileged framework for drug discovery. The structure-based drug discovery is the widely used method of drug repurposing, consisting of computational screening of compounds and testing them in-vitro. This current method of repurposing leaves room for mechanistic insights into how these screened hits interact with and influence their targets. We addressed this gap in the current study by integrating highly sensitive biophysical methods into existing computational repurposing methods. We also corroborated our computational and biophysical findings on H37Rv for the anti-mycobacterial action of selected drugs in-vitro and ex-vivo conditions. Atosiban and Rutin were screened as highly interacting hits against HemD through multi-stage docking and were cross-validated in biophysical studies. The affinity of these drugs (K ~ 106 M−1) was quantified using fluorescence quenching studies. Differential Scanning Fluorimetry (DSF) and urea-based chemical denaturation studies revealed a destabilizing effect of these drugs on target which was further validated using MD simulations. Conformational rearrangements of secondary structures were established using CD spectra and intrinsic fluorescence. Furthermore, Atosiban and Rutin inhibited M.tb growth in-vitro and ex-vivo while remaining non-toxic to mice peritoneal macrophages.

Accelerated screening of Tax(CoCrFeMnNi)1-x and Nbx(CoCrFeMnNi)1-x high-entropy alloys

High entropy alloys are a promising group of materials with seemingly limitless and unexplored composition spaces, capable of producing extraordinary physical/mechanical properties that exceed those of conventional materials. Utilizing an additive manufacturing enabled high throughput materials discovery framework, Ta and Nb graded additions to CoCrFeMnNi are explored. Compact metallurgical Ta,Nbx(CoCrFeMnNi)1-x specimens spanning x = 0 – 100 at.% were fabricated by powder based directed energy deposition. Microstructure evolution as a function of specific Ta,Nbx(CoCrFeMnNi)1-x compositions is investigated through a combination of XRD/XRF and SEM-EDS/EBSD. Microhardness measurements were also performed to rapidly establish baseline structure–property relationships. Results revealed that eutectic microstructures were largely suppressed throughout the explored composition ranges resulting in limited ductility and extraordinary hardness.

The Oncology Biomarker Discovery framework reveals cetuximab and bevacizumab response patterns in metastatic colorectal cancer

Precision medicine has revolutionised cancer treatments; however, actionable biomarkers remain scarce. To address this, we develop the Oncology Biomarker Discovery (OncoBird) framework for analysing the molecular and biomarker landscape of randomised controlled clinical trials. OncoBird identifies biomarkers based on single genes or mutually exclusive genetic alterations in isolation or in the context of tumour subtypes, and finally, assesses predictive components by their treatment interactions. Here, we utilise the open-label, randomised phase III trial (FIRE-3, AIO KRK-0306) in metastatic colorectal carcinoma patients, who received either cetuximab or bevacizumab in combination with 5-fluorouracil, folinic acid and irinotecan (FOLFIRI). We systematically identify five biomarkers with predictive components, e.g., patients with tumours that carry chr20q amplifications or lack mutually exclusive ERK signalling mutations benefited from cetuximab compared to bevacizumab. In summary, OncoBird characterises the molecular landscape and outlines actionable biomarkers, which generalises to any molecularly characterised randomised controlled trial.

Studying bias in visual features through the lens of optimal transport

Computer vision systems are employed in a variety of high-impact applications. However, making them trustworthy requires methods for the detection of potential biases in their training data, before models learn to harm already disadvantaged groups in downstream applications. Image data are typically represented via extracted features, which can be hand-crafted or pre-trained neural network embeddings. In this work, we introduce a framework for bias discovery given such features that is based on optimal transport theory; it uses the (quadratic) Wasserstein distance to quantify disparity between the feature distributions of two demographic groups (e.g., women vs men). In this context, we show that the Kantorovich potentials of the images, which are a byproduct of computing the Wasserstein distance and act as “transportation prices", can serve as bias scores by indicating which images might exhibit distinct biased characteristics. We thus introduce a visual dataset exploration pipeline that helps auditors identify common characteristics across high- or low-scored images as potential sources of bias. We conduct a case study to identify prospective gender biases and demonstrate theoretically-derived properties with experiments on the CelebA and Biased MNIST datasets.

SocialSift: Target Query Discovery on Online Social Media With Deep Reinforcement Learning.

Among the prohibitively large volume of posts (e.g., tweets in Twitter) on online social networks (OSNs), how to design effective queries to explore the ones of interest is a pressing problem. There are two main challenges to address the problem. First, given public application programming interfaces (APIs) for querying posts related to keywords from an extremely large vocabulary, how to infer the keywords relevant to our target interest using as few queries as possible? Second, how to deal with the agnostics of OSN's API? i.e., as different social networks typically have different running mechanisms, even with some randomness in returning results, how to build the knowledge of the API returns w.r.t. target interests from scratches? To address the above two challenges, we propose a target query discovery framework based on a deep reinforcement learning approach, named SocialSift. SocialSift intelligently interacts with OSNs' keyword-based API and develops its own knowledge in searching the optimal queries w.r.t. the target interests as well as OSN APIs. Specifically, to address the first challenge, we are inspired by the human searching experience, and recognize learning to query with context awareness to reduce the searching space, by qualifying keywords from returned results and keeping the tracks of the query trial history, or say contexts. As for addressing the second challenge, we treat OSNs' APIs as black boxes and probabilistically quantify query-interest pairs guided by rewards, which is a well-curated indicator w.r.t. target interests. Empirical results on three popular OSNs: Twitter, Reddit, and Amazon demonstrate our SocialSift significantly outperforms the state-of-the-art baselines by 12% in retrieving target posts.

A framework for REST services discovery and composition

A framework for REST services discovery and composition

Robust model agnostic predictive control algorithm for randomly excited dynamical systems

We propose a novel robust model agnostic predictive control (RoMAn-MPC) algorithm and illustrate its application in randomly excited dynamical systems. Unlike conventional model predictive control algorithms, the proposed RoMAn-MPC requires no information about the underlying physics of the system; instead, the governing physics is identified by using the recently proposed stochastic equation discovery framework. One key advantage of the proposed approach resides in its capability to generalize; this eliminates the repeated retraining phase — a major bottleneck with other machine learning-based model agnostic control algorithms. Overall, the proposed RoMAn-MPC (a) is robust against measurement noise, (b) works with sparse measurements, (c) can tackle set-point changes, (d) works with multiple control variables, and (e) can incorporate dead time. We have obtained state-of-the-art results on several benchmark examples. Finally, we use the proposed approach for vibration mitigation of a 76-storey building under seismic loading.

Voxelized atomic structure framework for materials design and discovery

We present a computational framework for developing physics-based, high-fidelity structure–property relationships with atomic systems. In this framework, atomic structure is quantified by directionally resolved two-point spatial correlations of the charge density field, projected to a salient low-dimensional feature space via principal component analysis (PCA), and correlated to physical properties by Gaussian process regression (GPR). The charge density field provides a complete, purely physics-based definition of the atomic structure that is independent of chemical species information and does not require additional feature engineering or idealizations beyond those of first-principles computations. The two-point spatial correlations capture the salient spatial features underlying the atomic structure that dictate the physics underlying the material response. Since the feature engineering approach explored in this work is universally applicable to all atomic structures independent of the chemical species present in the structure, it offers new avenues for efficiently exploring the space of atomic structures for desired property combinations. A further contribution of this work comes from utilizing the uncertainty quantification inherently provided by GPR to deploy a Bayesian experiment design strategy to minimize the number of computationally expensive physics simulations required to achieve the desired accuracy. In this work, we demonstrate the proposed framework to elucidate the relationship between the chemical composition and bulk modulus in AlNbTiZr high entropy alloys. It is shown that a highly accurate structure–property relationship with less than 2% average error can be established using a small training dataset of less than 30 samples.

Brain effective connectome based on fMRI and DTI data: Bayesian causal learning and assessment.

Neuroscientific studies aim to find an accurate and reliable brain Effective Connectome (EC). Although current EC discovery methods have contributed to our understanding of brain organization, their performances are severely constrained by the short sample size and poor temporal resolution of fMRI data, and high dimensionality of the brain connectome. By leveraging the DTI data as prior knowledge, we introduce two Bayesian causal discovery frameworks -the Bayesian GOLEM (BGOLEM) and Bayesian FGES (BFGES) methods- that offer significantly more accurate and reliable ECs and address the shortcomings of the existing causal discovery methods in discovering ECs based on only fMRI data. Moreover, to numerically assess the improvement in the accuracy of ECs with our method on empirical data, we introduce the Pseudo False Discovery Rate (PFDR) as a new computational accuracy metric for causal discovery in the brain. Through a series of simulation studies on synthetic and hybrid data (combining DTI from the Human Connectome Project (HCP) subjects and synthetic fMRI), we demonstrate the effectiveness of our proposed methods and the reliability of the introduced metric in discovering ECs. By employing the PFDR metric, we show that our Bayesian methods lead to significantly more accurate results compared to the traditional methods when applied to the Human Connectome Project (HCP) data. Additionally, we measure the reproducibility of discovered ECs using the Rogers-Tanimoto index for test-retest data and show that our Bayesian methods provide significantly more reliable ECs than traditional methods. Overall, our study's numerical and visual results highlight the potential for these frameworks to significantly advance our understanding of brain functionality.

A novel Hyperledger blockchain-enabled decentralized application for drug discovery chain management

The aim of the drug discovery process is to create a brand-new drug that is both safe and useful for treating disease in patients. The trustworthy collaboration of the organisations involved in the drug discovery process and the integrity of their contributions are essential for this process. Current drug discovery chains use centralised systems, which are susceptible to lockdown by cyberattacks. Blockchain, with its many characteristics such as accountability, immutability, integrity, privacy, and security, has the potential to be extremely useful in drug discovery chain management. The purpose of this work is to create a drug discovery framework utilizing the strengths of blockchain technology in combination with machine learning (ML). Such a system would provide a secure, efficient, and faster drug development life cycle. This study presents a novel Hyperledger fabric-based drug discovery application that empowers the permissioned organisations to upload, update, view, and verify contributions. ML is used to preprocess data and visualise features. In the proposed work, a unique identifier is assigned to each contribution asset using the secure hash algorithm (SHA-256). The proposed design also enables the regulatory authority to issue certificates proving the ownership of contributions to the contributing organisations. The blockchain ledger has been used to store the meta-data and InChIKey of drug contributions, and actual contributions are in off-chain storage. In this work, we have successfully built an end-to-end decentralised drug discovery application with a front-end interface and demonstrated chaincode algorithms. The end-to-end application is not available in any previous work. The Caliper tool has been used to investigate throughput, latency, and resource statistics. The performance and comparative analysis show that the proposed design is scalable and promising for drug discovery chain management.

CellDeathPred: a deep learning framework for ferroptosis and apoptosis prediction based on cell painting

Cell death, such as apoptosis and ferroptosis, play essential roles in the process of development, homeostasis, and pathogenesis of acute and chronic diseases. The increasing number of studies investigating cell death types in various diseases, particularly cancer and degenerative diseases, has raised hopes for their modulation in disease therapies. However, identifying the presence of a particular cell death type is not an obvious task, as it requires computationally intensive work and costly experimental assays. To address this challenge, we present CellDeathPred, a novel deep-learning framework that uses high-content imaging based on cell painting to distinguish cells undergoing ferroptosis or apoptosis from healthy cells. In particular, we incorporate a deep neural network that effectively embeds microscopic images into a representative and discriminative latent space, classifies the learned embedding into cell death modalities, and optimizes the whole learning using the supervised contrastive loss function. We assessed the efficacy of the proposed framework using cell painting microscopy data sets from human HT-1080 cells, where multiple inducers of ferroptosis and apoptosis were used to trigger cell death. Our model confidently separates ferroptotic and apoptotic cells from healthy controls, with an average accuracy of 95% on non-confocal data sets, supporting the capacity of the CellDeathPred framework for cell death discovery.

Network embedding framework for driver gene discovery by combining functional and structural information

Comprehensive analysis of multiple data sets can identify potential driver genes for various cancers. In recent years, driver gene discovery based on massive mutation data and gene interaction networks has attracted increasing attention, but there is still a need to explore combining functional and structural information of genes in protein interaction networks to identify driver genes. Therefore, we propose a network embedding framework combining functional and structural information to identify driver genes. Firstly, we combine the mutation data and gene interaction networks to construct mutation integration network using network propagation algorithm. Secondly, the struc2vec model is used for extracting gene features from the mutation integration network, which contains both gene's functional and structural information. Finally, machine learning algorithms are utilized to identify the driver genes. Compared with the previous four excellent methods, our method can find gene pairs that are distant from each other through structural similarities and has better performance in identifying driver genes for 12 cancers in the cancer genome atlas. At the same time, we also conduct a comparative analysis of three gene interaction networks, three gene standard sets, and five machine learning algorithms. Our framework provides a new perspective for feature selection to identify novel driver genes.