Enhancing molecular property prediction with auxiliary learning and task-specific adaptation.

Pretrained Graph Neural Networks have been widely adopted for various molecular property prediction tasks. Despite their ability to encode structural and relational features of molecules, traditional fine-tuning of such pretrained GNNs on the target task can lead to poor generalization. To address this, we explore the adaptation of pretrained GNNs to the target task by jointly training them with multiple auxiliary tasks. This could enable the GNNs to learn both general and task-specific features, which may benefit the target task. However, a major challenge is to determine the relatedness of auxiliary tasks with the target task. To address this, we investigate multiple strategies to measure the relevance of auxiliary tasks and integrate such tasks by adaptively combining task gradients or by learning task weights via bi-level optimization. Additionally, we propose a novel gradient surgery-based approach, Rotation of Conflicting Gradients ( RCGrad ), that learns to align conflicting auxiliary task gradients through rotation. Our experiments with state-of-the-art pretrained GNNs demonstrate the efficacy of our proposed methods, with improvements of up to 7.7% over fine-tuning. This suggests that incorporating auxiliary tasks along with target task fine-tuning can be an effective way to improve the generalizability of pretrained GNNs for molecular property prediction.Scientific contributionWe introduce a novel framework for adapting pretrained GNNs to molecular tasks using auxiliary learning to address the critical issue of negative transfer. Leveraging novel gradient surgery techniques such as RCGrad , the proposed adaptation framework represents a significant departure from the dominant pretraining fine-tuning approach for molecular GNNs. Our contributions are significant for drug discovery research, especially for tasks with limited data, filling a notable gap in the efficient adaptation of pretrained models for molecular GNNs.

Improving Anticancer Drug Selection and Prioritization via Neural Learning to Rank.

Personalized cancer treatment requires a thorough understanding of complex interactions between drugs and cancer cell lines in varying genetic and molecular contexts. To address this, high-throughput screening has been used to generate large-scale drug response data, facilitating data-driven computational models. Such models can capture complex drug-cell line interactions across various contexts in a fully data-driven manner. However, accurately prioritizing the most effective drugs for each cell line still remains a significant challenge. To address this, we developed multiple neural ranking approaches that leverage large-scale drug response data across multiple cell lines from diverse cancer types. Unlike existing approaches that primarily utilize regression and classification techniques for drug response prediction, we formulated the objective of drug selection and prioritization as a drug ranking problem. In this work, we proposed multiple pairwise and listwise neural ranking methods that learn latent representations of drugs and cell lines and then use those representations to score drugs in each cell line via a learnable scoring function. Specifically, we developed neural pairwise and listwise ranking methods, Pair-PushC and List-One on top of the existing methods, pLETORg and ListNet, respectively. Additionally, we proposed a novel listwise ranking method, List-All, that focuses on all the effective drugs instead of the top effective drug, unlike List-One. We also provide an exhaustive empirical evaluation with state-of-the-art regression and ranking baselines on large-scale data sets across multiple experimental settings. Our results demonstrate that our proposed ranking methods mostly outperform the best baselines with significant improvements of as much as 25.6% in terms of selecting truly effective drugs within the top 20 predicted drugs (i.e., hit@20) across 50% test cell lines. Furthermore, our analyses suggest that the learned latent spaces from our proposed methods demonstrate informative clustering structures and capture relevant underlying biological features. Moreover, our comprehensive evaluation provides a thorough and objective comparison of the performance of different methods (including our proposed ones).

The relationship between histopathological features, immunosuppression and outcome in patients undergoing native kidney biopsies.

AIMS: To assess retrospectively the association between histopathological lesions on renal biopsy and subsequent impairment of renal function across the spectrum of kidney diseases and to explore the influence of immunosuppressive therapy within the first 6 months after biopsy on this association. METHODS AND RESULTS: Clinical data from 488 adult patients having a renal biopsy reported at a single centre from 2017 to 2019 were obtained during a median follow-up period of 786 days. Seventeen semi-quantitative histology parameters were recorded at the time of biopsy, 14 of which were suitable for assessment of association with loss of eGFR by multivariable Cox regression analysis, measurement of eGFR slope and measurement of eGFR 12 months after biopsy. A widely used histopathological chronicity score was also assessed. Clinical baseline variables including prescription of immunosuppression were recorded. Seven of 14 histology parameters: mesangial matrix expansion, global glomerulosclerosis, tubular atrophy, interstitial fibrosis, arteriolosclerosis, mesangial hypercellularity and acute tubular injury; and the chronicity score, predicted loss of kidney function by all three measures. Prescription of immunosuppression was more likely in patients with active inflammatory pathology and less likely in patients with chronic fibrotic pathology, and was associated with reduced risk of loss of eGFR. CONCLUSIONS: This retrospective study demonstrates the prognostic significance and complex relationship with immunosuppression of routinely reported histopathological variables in patients having native kidney biopsies, across the spectrum of kidney diseases. It provides useful information for renal biopsy prognostication and design of retrospective studies, including machine learning models.

Outcomes in ANCA-Associated Vasculitis in Scotland: Validation of the Renal Risk Score in a Complete National Cohort.

Introduction: Antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) causes autoimmune-mediated inflammation of small blood vessels in multiple organs, including the kidneys. The ability to accurately predict kidney outcomes would enable a more personalized therapeutic approach. Methods: We used our national renal biopsy registry to validate the ability of ANCA Renal Risk Score (ARRS) to predict end-stage kidney disease (ESKD) for individual patients. This score uses histopathological and biochemical data to stratify patients as high, medium, or low risk for developing ESKD. Results: A total of 288 patients were eligible for inclusion in the study (low risk n = 144, medium risk n = 122, high risk n = 12). Using adjusted Cox proportional hazard models with the low-risk group as reference, we show that outcome differs between the categories: high-risk hazard ratio (HR) 16.69 (2.91-95.81, P = 0.002); medium risk HR 4.14 (1.07-16.01, P = 0.039). Incremental multivariable-adjusted Cox proportional hazards models demonstrated that adding ARRS to a model adjusted for multiple clinical parameters enhanced predictive discrimination (basic model C-statistic 0.864 [95% CI 0.813-0.914], basic model plus ARRS C-statistic 0.877 [95% CI 0.823-0.931]; P <0.01). Conclusion: The ARRS better discriminates risk of ESKD in AAV and offers clinicians more prognostic information than the use of standard biochemical and clinical measures alone. This is the first time the ARRS has been validated in a national cohort. The proportion of patients with high-risk scores is lower in our cohort compared to others and should be noted as a limitation of this study.

A Pipeline to Understand Emerging Illness Via Social Media Data Analysis: Case Study on Breast Implant Illness.

BACKGROUND: A new illness can come to public attention through social media before it is medically defined, formally documented, or systematically studied. One example is a condition known as breast implant illness (BII), which has been extensively discussed on social media, although it is vaguely defined in the medical literature. OBJECTIVE: The objective of this study is to construct a data analysis pipeline to understand emerging illnesses using social media data and to apply the pipeline to understand the key attributes of BII. METHODS: We constructed a pipeline of social media data analysis using natural language processing and topic modeling. Mentions related to signs, symptoms, diseases, disorders, and medical procedures were extracted from social media data using the clinical Text Analysis and Knowledge Extraction System. We mapped the mentions to standard medical concepts and then summarized these mapped concepts as topics using latent Dirichlet allocation. Finally, we applied this pipeline to understand BII from several BII-dedicated social media sites. RESULTS: Our pipeline identified topics related to toxicity, cancer, and mental health issues that were highly associated with BII. Our pipeline also showed that cancers, autoimmune disorders, and mental health problems were emerging concerns associated with breast implants, based on social media discussions. Furthermore, the pipeline identified mentions such as rupture, infection, pain, and fatigue as common self-reported issues among the public, as well as concerns about toxicity from silicone implants. CONCLUSIONS: Our study could inspire future studies on the suggested symptoms and factors of BII. Our study provides the first analysis and derived knowledge of BII from social media using natural language processing techniques and demonstrates the potential of using social media information to better understand similar emerging illnesses.

ANCA-associated renal vasculitis is associated with rurality but not seasonality or deprivation in a complete national cohort study.

BACKGROUND: Small studies suggest an association between ANCA-associated vasculitis (AAV) incidence and rurality, seasonality and socioeconomic deprivation. We examined the incidence of kidney biopsy-proven AAV and its relationship with these factors in the adult Scottish population. METHODS: Using the Scottish Renal Biopsy Registry, all adult native kidney biopsies performed between 2014 and 2018 with a diagnosis of granulomatosis with polyangiitis (GPA) or microscopic polyangiitis (MPA) were identified. The Scottish Government Urban Rural Classification was used for rurality analysis. Seasons were defined as autumn (September-November), winter (December-February), spring (March-May) and summer (June-August). Patients were separated into quintiles of socioeconomic deprivation using the validated Scottish Index of Multiple Deprivation and incidence standardised to age. Estimated glomerular filtration rate and urine protein:creatinine ratio at time of biopsy were used to assess disease severity. RESULTS: 339 cases of renal AAV were identified, of which 62% had MPA and 38% had GPA diagnosis. AAV incidence was 15.1 per million population per year (pmp/year). Mean age was 66 years and 54% were female. Incidence of GPA (but not MPA) was positively associated with rurality (5.2, 8.4 and 9.1 pmp/year in 'urban', 'accessible remote' and 'rural remote' areas, respectively; p=0.04). The age-standardised incidence ratio was similar across all quintiles of deprivation (p=ns). CONCLUSIONS: Seasonality and disease severity did not vary across AAV study groups. In this complete national cohort study, we observed a positive association between kidney biopsy-proven GPA and rurality.

A National Registry Study of Patient and Renal Survival in Adult Nephrotic Syndrome.

INTRODUCTION: We aimed to determine the mortality rate, cause of death, and rate of end-stage kidney disease (ESKD) in adults with nephrotic syndrome (NS). METHODS: We conducted a national registry-based study, including all 522 adults who had a kidney biopsy for NS in Scotland in 2014-2017. We linked the Scottish Renal Registry to death certificate data. We performed survival and Cox proportional hazards analyses, accounting for competing risks of death and ESKD. We compared mortality rates with those in the age- and sex-matched general population. RESULTS: A total of 372 patients had primary NS; 150 had secondary NS. Over a median follow-up of 866 days, 110 patients (21%) died. In patients with primary NS, observed versus population 3-year mortality was 2.1% (95% CI 0.0%-4.6%) versus 0.9% (0.8%-1.0%) in patients aged <60 years and 24.9% (18.4%-30.8%) versus 9.4% (8.3%-10.5%) in those aged ≥60 years. In secondary NS, this discrepancy was 17.1% (5.6%-27.2%) versus 1.1% (0.9%-1.2%) in <60-year-olds and 49.4% (36.6%-59.7%) versus 8.1% (6.6%-9.6%) in ≥60-year-olds. In primary NS, cardiovascular causes accounted for 28% of deaths, compared with 18% in the general population. Eighty patients (15%) progressed to ESKD. Incidence of ESKD by 3 years was 8.4% (95% CI 4.9%-11.7%) in primary and 35.1% (24.3%-44.5%) in secondary NS. Early remission of proteinuria and the absence of early acute kidney injury (AKI) were associated with lower rates of death and ESKD. CONCLUSIONS: Adults with NS have high rates of death and ESKD. Cardiovascular causes account for excess mortality in primary NS.