Drug resistance is a major obstacle to curative cancer therapies, and increased understanding of the molecular events contributing to resistance would enable better prediction of therapy response, as well as contribute to new targets for combination therapy. Here we have analyzed the early molecular response to epidermal growth factor receptor (EGFR) inhibition using RNA sequencing data covering 13 486 genes and mass spectrometry data covering 10 138 proteins. This analysis revealed a massive response to EGFR inhibition already within the first 24 hours, including significant regulation of hundreds of genes known to control downstream signaling, such as transcription factors, kinases, phosphatases and ubiquitin E3-ligases. Importantly, this response included upregulation of key genes in multiple oncogenic signaling pathways that promote proliferation and survival, such as ERBB3, FGFR2, JAK3 and BCL6, indicating an early adaptive response to EGFR inhibition. Using a library of more than 500 approved and experimental compounds in a combination therapy screen, we could show that several kinase inhibitors with targets including JAK3 and FGFR2 increased the response to EGFR inhibitors. Further, we investigated the functional impact of BCL6 upregulation in response to EGFR inhibition using siRNA-based silencing of BCL6. Proteomics profiling revealed that BCL6 inhibited transcription of multiple target genes including p53, resulting in reduced apoptosis which implicates BCL6 upregulation as a new EGFR inhibitor treatment escape mechanism. Finally, we demonstrate that combined treatment targeting both EGFR and BCL6 act synergistically in killing lung cancer cells. In conclusion, or data indicates that multiple different adaptive mechanisms may act in concert to blunt the cellular impact of EGFR inhibition, and we suggest BCL6 as a potential target for EGFR inhibitor-based combination therapy.

The onset of obesity-linked type 2 diabetes (T2D) is marked by an eventual failure in pancreatic β-cell function and mass that is no longer able to compensate for the inherent insulin resistance and increased metabolic load intrinsic to obesity. However, in a commonly used model of T2D, the db/db mouse, β-cells have an inbuilt adaptive flexibility enabling them to effectively adjust insulin production rates relative to the metabolic demand. Pancreatic β-cells from these animals have markedly reduced intracellular insulin stores, yet high rates of (pro)insulin secretion, together with a substantial increase in proinsulin biosynthesis highlighted by expanded rough endoplasmic reticulum and Golgi apparatus. However, when the metabolic overload and/or hyperglycemia is normalized, β-cells from db/db mice quickly restore their insulin stores and normalize secretory function. This demonstrates the β-cell’s adaptive flexibility and indicates that therapeutic approaches applied to encourage β-cell rest are capable of restoring endogenous β-cell function. However, mechanisms that regulate β-cell adaptive flexibility are essentially unknown. To gain deeper mechanistic insight into the molecular events underlying β-cell adaptive flexibility in db/db β-cells, we conducted a combined proteomic and post-translational modification specific proteomic (PTMomics) approach on islets from db/db mice and wild-type controls (WT) with or without prior exposure to normal glucose levels. We identified differential modifications of proteins involved in redox homeostasis, protein refolding, K48-linked deubiquitination, mRNA/protein export, focal adhesion, ERK1/2 signaling, and renin-angiotensin-aldosterone signaling, as well as sialyltransferase activity, associated with β-cell adaptive flexibility. These proteins are all related to proinsulin biosynthesis and processing, maturation of insulin secretory granules, and vesicular trafficking—core pathways involved in the adaptation of insulin production to meet metabolic demand. Collectively, this study outlines a novel and comprehensive global PTMome signaling map that highlights important molecular mechanisms related to the adaptive flexibility of β-cell function, providing improved insight into disease pathogenesis of T2D.

Post-translational modifications play a critical and diverse role in regulating cellular activities. Despite their fundamentally important role in cellular function, there has been no report to date of an effective generalized approach to the targeting, extraction and characterization of the critical c-terminal regions of natively prenylated proteins. Various chemical modification and metabolic labelling strategies in cell culture have been reported. However, their applicability is limited to cell culture systems and does not allow for analysis of tissue samples. The chemical characteristics (hydrophobicity, low abundance, highly basic charge) of many of the c-terminal regions of prenylated proteins have impaired the use of standard proteomic workflows. In this context, we sought a direct approach to the problem in order to examine these proteins in tissue without the use of labelling.  Here we demonstrate that prenylated proteins can be captured on chromatographic resins functionalized with mixed disulfide functions. Protease treatment of resin-bound proteins using chymotryptic digestion revealed peptides from many known prenylated proteins. Exposure of the protease-treated resin to reducing agents and hydro organic mixtures released c-terminal peptides with intact prenyl groups along with other enzymatic modifications expected in this protein family. Database and search parameters were selected to allow for c-terminal modifications unique to these molecules such as CAAX box processing and c-terminal methylation. In summary, we present a direct approach to enrich and obtain information at a molecular level of detail about prenylation of proteins from tissue and cell extracts using high performance LCMS without the need for metabolic labeling and derivatization.

HLA-DQA1 and -DQB1 are strongly associated with type 1 diabetes (T1D), and DQ8.1 and DQ2.5 are major risk haplotypes. Next generation targeted sequencing of HLA-DQA1 and -DQB1 in Swedish newly diagnosed 1-18 year-old patients (n=962) and controls (n=636) was used to construct abbreviated DQ haplotypes, converted into amino acid (AA) residues, and assessed for their associations with T1D. A hierarchically-organized haplotype (HOH) association analysis, allowed 45 unique DQ haplotypes to be categorized into seven clusters. The DQ8/9 cluster included two DQ8.1 risk and the DQ9 resistant haplotypes, and the DQ2 cluster, included the DQ2.5 risk and DQ2.2 resistant haplotypes. Within each cluster, HOH found residues α44Q (OR 3.29, p=2.38*10^-85 ) and β57A (OR 3.44, p=3.80*10^-84) to be associated with T1D in the DQ8/9 cluster representing all ten residues (α22, α23, α44, α49, α51, α53, α54, α73, α184, β57) due to complete linkage-disequilibrium (LD) of α44 with eight such residues. Within the DQ2 cluster and due to LD, HOH analysis found α44C and β135D to share the risk for T1D (OR 2.10, p=1.96*10^-20). The motif "QAD" of α44, β57, and β135 captured the T1D risk association of DQ8.1 (OR 3.44, p=3.80*10^-84), the corresponding motif "CAD" captured the risk association of DQ2.5 (OR 2.10, p=1.96*10^-20). Two risk associations were related to GADA and IA-2A, but in opposite directions. "CAD" was positively associated with GADA (OR 1.56; p=6.35*10^-8) but negatively with IA-2A (OR 0.59, p= 6.55*10^-11). "QAD" was negatively associated with GADA (OR 0.88; p= 3.70*10^-3) but positively with IA-2A (OR 1.64; p= 2.40*10^-14), despite a single difference at α44. The residues are found in and around anchor pockets 1 and 9, as potential TCR contacts, in the areas for CD4 binding and putative homodimer formation. The identification of three HLA-DQ AA (α44, β57, β135) conferring T1D risk should sharpen functional and translational studies.

Paraphrasing the Swiss physician and father of toxicology Paracelsus (1493–1541) on chemical agents used as therapeutics, "the dose makes the poison," it is now realized that this aptly applies to the calorigenic nutrients. The case here is the pancreatic islet β-cell presented with excessive levels of nutrients such as glucose, lipids, and amino acids. The short-term effects these nutrients exert on the β-cell are enhanced insulin biosynthesis and secretion and changes in glucose sensitivity. However, chronic fuel surfeit triggers additional compensatory and adaptive mechanisms by β-cells to cope with the increased insulin demand or to protect itself. When these mechanisms fail, toxicity due to the nutrient surplus ensues, leading to β-cell dysfunction, dedifferentiation, and apoptosis. The terms glucotoxicity, lipotoxicity, and glucolipotoxicity have been widely used, but there is some confusion as to what they mean precisely and which is most appropriate for a given situation. Here we address the gluco-, lipo-, and glucolipo-toxicities in β-cells by assessing the evidence both for and against each of them. We also discuss potential mechanisms and defend the view that many of the identified "toxic" effects of nutrient excess, which may also include amino acids, are in fact beneficial adaptive processes. In addition, candidate fuel-excess detoxification pathways are evaluated. Finally, we propose that a more general term should be used for the in vivo situation of overweight-associated type 2 diabetes reflecting both the adaptive and toxic processes to mixed calorigenic nutrients excess: "nutrient-induced metabolic stress" or, in brief, "nutri-stress."

Motor protein-based active transport is essential for mRNA localization and local translation in animal cells, yet how mRNA granules interact with motor proteins remains poorly understood. Using an unbiased yeast two–hybrid screen for interactions between murine RNA-binding proteins (RBPs) and motor proteins, here we identified protein interaction with APP tail-1 (PAT1) as a potential direct adapter between zipcode-binding protein 1 (ZBP1, a β-actin RBP) and the kinesin-I motor complex. The amino acid sequence of mouse PAT1 is similar to that of the kinesin light chain (KLC), and we found that PAT1 binds to KLC directly. Studying PAT1 in mouse primary hippocampal neuronal cultures from both sexes and using structured illumination microscopic imaging of these neurons, we observed that brain-derived neurotrophic factor (BDNF) enhances co-localization of dendritic ZBP1 and PAT1 within granules that also contain kinesin-I. PAT1 is essential for BDNF-stimulated neuronal growth cone development and dendritic protrusion formation, and we noted that ZBP1 and PAT1 co-locate along with β-actin mRNA in actively transported granules in living neurons. Acute disruption of the PAT1–ZBP1 interaction in neurons with PAT1 siRNA or a dominant-negative ZBP1 construct diminished localization of β-actin mRNA but not of Ca2+/calmodulin-dependent protein kinase IIα (CaMKIIα) mRNA in dendrites. The aberrant β-actin mRNA localization resulted in abnormal dendritic protrusions and growth cone dynamics. These results suggest a critical role for PAT1 in BDNF-induced β-actin mRNA transport during postnatal development and reveal a new molecular mechanism for mRNA localization in vertebrates.

12 June 2019

Research Event

Event participants

Dr Jon Gibbins, Senior Lecturer in the Department of Mechanical Engineering, Imperial College London
Jim Footner, Senior Climate Change Campaigner, Greenpeace

Aging-dependent changes in tissue function are associated with the development of metabolic diseases. However, the molecular connections linking aging, obesity, and diabetes remain unclear. Lamin A, lamin C, and progerin, products of the Lmna gene, have antagonistic functions on energy metabolism and life span. Lamin C, albeit promoting obesity, increases life span, suggesting that this isoform is crucial for maintaining healthy conditions under metabolic stresses. Because β-cell loss during obesity or aging leads to diabetes, we investigated the contribution of lamin C to β-cell function in physiopathological conditions. We demonstrate that aged lamin C only–expressing mice (Lmna^LCS/LCS) become obese but remain glucose tolerant due to adaptive mechanisms including increased β-cell mass and insulin secretion. Triggering diabetes in young mice revealed that Lmna^LCS/LCS animals normalize their fasting glycemia by both increasing insulin secretion and regenerating β-cells. Genome-wide analyses combined to functional analyses revealed an increase of mitochondrial biogenesis and global translational rate in Lmna^LCS/LCS islets, two major processes involved in insulin secretion. Altogether, our results demonstrate for the first time that the sole expression of lamin C protects from glucose intolerance through a β-cell–adaptive transcriptional program during metabolic stresses, highlighting Lmna gene processing as a new therapeutic target for diabetes treatment.

The European Medicines Agency (EMA) has embraced a new model of drug testing and marketing called “adaptive pathways”, allowing new drugs for “unmet medical needs” to be launched on the market faster, on the basis of fewer data. While industry claims this is necessary, an analysis on thebmj.com looks at the assumptions underlying the new pathway,...

14 November 2019

Invitation Only Research Event

Event participants

Richard King, Senior Research Fellow, Energy, Environment and Resources Department, Chatham House
Chair: Duncan Brack, Associate Fellow, Energy, Environment and Resources Department, Chatham House

29 January 2020

Policymakers are in danger of sleepwalking into ineffective carbon dioxide removal solutions in the quest to tackle climate change. This paper warns against overreliance on bioenergy with carbon capture and storage (BECCS). 

Although long one of the world's top migrant destinations, only in the recent past has Germany come to acknowledge and adjust to its role as a country of immigration. Its welcoming approach—a relatively new development—has been put to the test amid massive humanitarian inflows beginning in 2015. This country profile examines Germany's history on immigration and highlights current and emerging debates.

OBJECTIVE

Haptoglobin is an acute-phase reactant with pleiotropic functions. We aimed to study whether urine haptoglobin may predict risk of mortality in people with type 2 diabetes.

The American Dental Association sent a letter to third-party payers urging that administrators of dental benefit plans adjust and adapt reimbursement procedures important to dentists and patients — including coverage for temporary procedures and adjusting fee schedules to account for cost of increasing infection control procedures ¬— in the midst of the “unprecedented and extraordinary circumstances dentists and their patients face” during the pandemic.

While the coordinated attacks of Velociraptor dinosaurs depicted in the 1993 blockbuster made for compelling movie viewing, a study published this week claims raptors most likely hunted solo, not in packs.

New research suggests flies that are unable to fly sleep more as they learn to adapt to their disability.

The Navy has been particularly hard hit by the coronavirus pandemic and is working to adapt its strategies for maintenance as well as containing outbreaks on ships.

"It is timely that the higher education sector examines paths forward to address and indeed engage in the new environment in which it will work in the future. This fellowship explored two approaches to engage students more deeply in their education. The first approach is generally termed 'interdisciplinary studies' where students define their own program of study; the second approach focuses on topic structure where students are given a wide range of choice and in effect can build a topic that suits their interests."--Page iv.

Krefft, Gerard, 1830-1881.

East Indians -- Australia -- Languages.

American essays -- 21st century.

Wurzburg : Stahel, 1867.

Berlin : Hirschwald, 1887.

London : Rebman, 1901.

London : Longmans, Green, 1872.

[London] : Chapman and Hall, [1877]

London : Macmillan, 1897.

London : Macmillan, 1886.

London : Bailliere, Tindall and Cox, 1884.

In Tuesday night's Republican primary in West Virginia, Robert Karnes, a West Virginia Republican state senator who lashed out at teachers during their nine-day strike, lost to pro-labor candidate Bill Hamilton.

[London?]

[London?], [between 1700 and 1799?]

Entrepreneur and business founder Elon Musk will donate $423,000 for laptops in the Michigan district, a few months after making a gift focused on improving water quality in the school system.

Heiko Werner, Hajo Holzmann, Pierre Vandekerkhove.

Source: Electronic Journal of Statistics, Volume 14, Number 1, 1816--1871.

Abstract:
We investigate a flexible two-component semiparametric mixture of regressions model, in which one of the conditional component distributions of the response given the covariate is unknown but assumed symmetric about a location parameter, while the other is specified up to a scale parameter. The location and scale parameters together with the proportion are allowed to depend nonparametrically on covariates. After settling identifiability, we provide local M-estimators for these parameters which converge in the sup-norm at the optimal rates over Hölder-smoothness classes. We also introduce an adaptive version of the estimators based on the Lepski-method. Sup-norm bounds show that the local M-estimator properly estimates the functions globally, and are the first step in the construction of useful inferential tools such as confidence bands. In our analysis we develop general results about rates of convergence in the sup-norm as well as adaptive estimation of local M-estimators which might be of some independent interest, and which can also be applied in various other settings. We investigate the finite-sample behaviour of our method in a simulation study, and give an illustration to a real data set from bioinformatics.

Gwenaëlle Castellan, Anthony Cousien, Viet Chi Tran.

Source: Electronic Journal of Statistics, Volume 14, Number 1, 50--81.

Abstract:
Global sensitivity analysis is a set of methods aiming at quantifying the contribution of an uncertain input parameter of the model (or combination of parameters) on the variability of the response. We consider here the estimation of the Sobol indices of order 1 which are commonly-used indicators based on a decomposition of the output’s variance. In a deterministic framework, when the same inputs always give the same outputs, these indices are usually estimated by replicated simulations of the model. In a stochastic framework, when the response given a set of input parameters is not unique due to randomness in the model, metamodels are often used to approximate the mean and dispersion of the response by deterministic functions. We propose a new non-parametric estimator without the need of defining a metamodel to estimate the Sobol indices of order 1. The estimator is based on warped wavelets and is adaptive in the regularity of the model. The convergence of the mean square error to zero, when the number of simulations of the model tend to infinity, is computed and an elbow effect is shown, depending on the regularity of the model. Applications in Epidemiology are carried to illustrate the use of non-parametric estimators.

Uttam Bandyopadhyay, Shirsendu Mukherjee, Atanu Biswas.

Source: Brazilian Journal of Probability and Statistics, Volume 34, Number 2, 291--303.

Abstract:
In adaptive crossover design, our goal is to allocate more patients to a promising treatment sequence. The present work contains a very simple three period crossover design for two competing treatments where the allocation in period 3 is done on the basis of the data obtained from the first two periods. Assuming normality of response variables we use a reliability functional for the choice between two treatments. We calculate the allocation proportions and their standard errors corresponding to the possible treatment combinations. We also derive some asymptotic results and provide solutions on related inferential problems. Moreover, the proposed procedure is compared with a possible competitor. Finally, we use a data set to illustrate the applicability of the proposed design.

Salem M. Al-Gezeri, Robert G. Aykroyd.

Source: Brazilian Journal of Probability and Statistics, Volume 33, Number 3, 498--519.

Abstract:
The use of Markov random field (MRF) models has proven to be a fruitful approach in a wide range of image processing applications. It allows local texture information to be incorporated in a systematic and unified way and allows statistical inference theory to be applied giving rise to novel output summaries and enhanced image interpretation. A great advantage of such low-level approaches is that they lead to flexible models, which can be applied to a wide range of imaging problems without the need for significant modification. This paper proposes and explores the use of conditional MRF models for situations where multiple images are to be processed simultaneously, or where only a single image is to be reconstructed and a sequential approach is taken. Although the coupling of image intensity values is a special case of our approach, the main extension over previous proposals is to allow the direct coupling of other properties, such as smoothness or texture. This is achieved using a local modulating function which adjusts the influence of global smoothing without the need for a fully inhomogeneous prior model. Several modulating functions are considered and a detailed simulation study, motivated by remote sensing applications in archaeological geophysics, of conditional reconstruction is presented. The results demonstrate that a substantial improvement in the quality of the image reconstruction, in terms of errors and residuals, can be achieved using this approach, especially at locations with rapid changes in the underlying intensity.

Umberto Amato, Anestis Antoniadis, Italia De Feis.

Source: Statistics Surveys, Volume 14, 32--70.

Abstract:
We present and compare some nonparametric estimation methods (wavelet and/or spline-based) designed to recover a one-dimensional piecewise-smooth regression function in both a fixed equidistant or not equidistant design regression model and a random design model. Wavelet methods are known to be very competitive in terms of denoising and compression, due to the simultaneous localization property of a function in time and frequency. However, boundary assumptions, such as periodicity or symmetry, generate bias and artificial wiggles which degrade overall accuracy. Simple methods have been proposed in the literature for reducing the bias at the boundaries. We introduce new ones based on adaptive combinations of two estimators. The underlying idea is to combine a highly accurate method for non-regular functions, e.g., wavelets, with one well behaved at boundaries, e.g., Splines or Local Polynomial. We provide some asymptotic optimal results supporting our approach. All the methods can handle data with a random design. We also sketch some generalization to the multidimensional setting. To study the performance of the proposed approaches we have conducted an extensive set of simulations on synthetic data. An interesting regression analysis of two real data applications using these procedures unambiguously demonstrates their effectiveness.

Oleksandr Sverdlov, Weng Kee Wong, Yevgen Ryeznik.

Source: Statistics Surveys, Volume 8, 2--44.

Abstract:
Adaptive clinical trials are becoming increasingly popular research designs for clinical investigation. Adaptive designs are particularly useful in phase I cancer studies where clinical data are scant and the goals are to assess the drug dose-toxicity profile and to determine the maximum tolerated dose while minimizing the number of study patients treated at suboptimal dose levels. In the current work we give an overview of adaptive design methods for phase I cancer trials. We find that modern statistical literature is replete with novel adaptive designs that have clearly defined objectives and established statistical properties, and are shown to outperform conventional dose finding methods such as the 3+3 design, both in terms of statistical efficiency and in terms of minimizing the number of patients treated at highly toxic or nonefficacious doses. We discuss statistical, logistical, and regulatory aspects of these designs and present some links to non-commercial statistical software for implementing these methods in practice.

Spatial trajectories are ubiquitous and complex signals. Their analysis is crucial in many research fields, from urban planning to neuroscience. Several approaches have been proposed to cluster trajectories. They rely on hand-crafted features, which struggle to capture the spatio-temporal complexity of the signal, or on Artificial Neural Networks (ANNs) which can be more efficient but less interpretable. In this paper we present a novel ANN architecture designed to capture the spatio-temporal patterns characteristic of a set of trajectories, while taking into account the demographics of the navigators. Hence, our model extracts markers linked to both behaviour and demographics. We propose a composite signal analyser (CompSNN) combining three simple ANN modules. Each of these modules uses different signal representations of the trajectory while remaining interpretable. Our CompSNN performs significantly better than its modules taken in isolation and allows to visualise which parts of the signal were most useful to discriminate the trajectories.

We focus on the challenge of finding a diverse collection of quality solutions on complex continuous domains. While quality diver-sity (QD) algorithms like Novelty Search with Local Competition (NSLC) and MAP-Elites are designed to generate a diverse range of solutions, these algorithms require a large number of evaluations for exploration of continuous spaces. Meanwhile, variants of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) are among the best-performing derivative-free optimizers in single-objective continuous domains. This paper proposes a new QD algorithm called Covariance Matrix Adaptation MAP-Elites (CMA-ME). Our new algorithm combines the self-adaptation techniques of CMA-ES with archiving and mapping techniques for maintaining diversity in QD. Results from experiments based on standard continuous optimization benchmarks show that CMA-ME finds better-quality solutions than MAP-Elites; similarly, results on the strategic game Hearthstone show that CMA-ME finds both a higher overall quality and broader diversity of strategies than both CMA-ES and MAP-Elites. Overall, CMA-ME more than doubles the performance of MAP-Elites using standard QD performance metrics. These results suggest that QD algorithms augmented by operators from state-of-the-art optimization algorithms can yield high-performing methods for simultaneously exploring and optimizing continuous search spaces, with significant applications to design, testing, and reinforcement learning among other domains.

Adaptive importance samplers are adaptive Monte Carlo algorithms to estimate expectations with respect to some target distribution which extit{adapt} themselves to obtain better estimators over a sequence of iterations. Although it is straightforward to show that they have the same $mathcal{O}(1/sqrt{N})$ convergence rate as standard importance samplers, where $N$ is the number of Monte Carlo samples, the behaviour of adaptive importance samplers over the number of iterations has been left relatively unexplored. In this work, we investigate an adaptation strategy based on convex optimisation which leads to a class of adaptive importance samplers termed extit{optimised adaptive importance samplers} (OAIS). These samplers rely on the iterative minimisation of the $chi^2$-divergence between an exponential-family proposal and the target. The analysed algorithms are closely related to the class of adaptive importance samplers which minimise the variance of the weight function. We first prove non-asymptotic error bounds for the mean squared errors (MSEs) of these algorithms, which explicitly depend on the number of iterations and the number of samples together. The non-asymptotic bounds derived in this paper imply that when the target belongs to the exponential family, the $L_2$ errors of the optimised samplers converge to the optimal rate of $mathcal{O}(1/sqrt{N})$ and the rate of convergence in the number of iterations are explicitly provided. When the target does not belong to the exponential family, the rate of convergence is the same but the asymptotic $L_2$ error increases by a factor $sqrt{ ho^star} > 1$, where $ ho^star - 1$ is the minimum $chi^2$-divergence between the target and an exponential-family proposal.

In many Machine Learning domains, datasets are characterized by highly imbalanced and overlapping classes. Particularly in the medical domain, a specific list of symptoms can be labeled as one of various different conditions. Some of these conditions may be more prevalent than others by several orders of magnitude. Here we present a novel unsupervised Domain Adaptation scheme for such datasets. The scheme, based on a specific type of Quantification, is designed to work under both label and conditional shifts. It is demonstrated on datasets generated from Electronic Health Records and provides high quality results for both Quantification and Domain Adaptation in very challenging scenarios. Potential benefits of using this scheme in the current COVID-19 outbreak, for estimation of prevalence and probability of infection, are discussed.

Machine learning models with both good predictability and high interpretability are crucial for decision support systems. Linear regression is one of the most interpretable prediction models. However, the linearity in a simple linear regression worsens its predictability. In this work, we introduce a locally adaptive interpretable regression (LoAIR). In LoAIR, a metamodel parameterized by neural networks predicts percentile of a Gaussian distribution for the regression coefficients for a rapid adaptation. Our experimental results on public benchmark datasets show that our model not only achieves comparable or better predictive performance than the other state-of-the-art baselines but also discovers some interesting relationships between input and target variables such as a parabolic relationship between CO2 emissions and Gross National Product (GNP). Therefore, LoAIR is a step towards bridging the gap between econometrics, statistics, and machine learning by improving the predictive ability of linear regression without depreciating its interpretability.

Reducing domain divergence is a key step in transfer learning problems. Existing works focus on the minimization of global domain divergence. However, two domains may consist of several shared subdomains, and differ from each other in each subdomain. In this paper, we take the local divergence of subdomains into account in transfer. Specifically, we propose to use low-dimensional manifold to represent subdomain, and align the local data distribution discrepancy in each manifold across domains. A Manifold Maximum Mean Discrepancy (M3D) is developed to measure the local distribution discrepancy in each manifold. We then propose a general framework, called Transfer with Manifolds Discrepancy Alignment (TMDA), to couple the discovery of data manifolds with the minimization of M3D. We instantiate TMDA in the subspace learning case considering both the linear and nonlinear mappings. We also instantiate TMDA in the deep learning framework. Extensive experimental studies demonstrate that TMDA is a promising method for various transfer learning tasks.

Effective property prediction methods can help accelerate the search for COVID-19 antivirals either through accurate in-silico screens or by effectively guiding on-going at-scale experimental efforts. However, existing prediction tools have limited ability to accommodate scarce or fragmented training data currently available. In this paper, we introduce a novel approach to learn predictors that can generalize or extrapolate beyond the heterogeneous data. Our method builds on and extends recently proposed invariant risk minimization, adaptively forcing the predictor to avoid nuisance variation. We achieve this by continually exercising and manipulating latent representations of molecules to highlight undesirable variation to the predictor. To test the method we use a combination of three data sources: SARS-CoV-2 antiviral screening data, molecular fragments that bind to SARS-CoV-2 main protease and large screening data for SARS-CoV-1. Our predictor outperforms state-of-the-art transfer learning methods by significant margin. We also report the top 20 predictions of our model on Broad drug repurposing hub.