clinical data analysts

Abstract: Gastroenteropancreatic neuroendocrine tumors (GEP-NET) lack prognostic biomarkers, and computed tomography (CT)-based body composition radiomics at the third lumbar vertebra (L3) is a promising candidate. Small cohorts make high-dimensional radiomic models prone to overfitting and single-split evaluations variable and optimistic. This methodological study proposes a four-step validation protocol for radiomic survival analysis in small cohorts: (1) false discovery rate (FDR)-corrected feature selection, (2) a dual-path comparison of prior feature selection versus embedded regularization, (3) repeated nested cross-validation (CV) and (4) a negative control with Gaussian noise features. In 78 GEP-NET patients, 535 radiomic features from five L3 compartments were used to evaluate five survival models with Harrell's concordance index (C-index), time-dependent area under the curve (AUC) and integrated Brier score (IBS). Step 1 selected at least one feature in 18% of outer folds under noise (nominal level 20%) versus 84% with radiomic features. The dual-path comparison showed no consistent difference between strategies. Repeated nested CV reduced variability across seeds (41% median reduction in C-index standard deviation), but not enough to distinguish models. The nested CV C-index exceeded chance by 0.07-0.15 with radiomic features but not clearly under noise (0.40-0.42 with prior feature selection, 0.47-0.55 with embedded regularization). Frequently selected features tentatively linked greater subcutaneous fat, skeletal muscle compactness and texture to better prognosis. These findings suggest a modest prognostic signal in L3 radiomics, with considerable uncertainty that the protocol quantifies. The protocol may transfer to other small-cohort radiomic survival studies. Code: https://github.com/marielatalaveram/radiomics-survival-validation-protocol.

Abstract: Missing ratings can overturn a personalized model's current prediction. For completed models with additive response scores, exact worst case bounds check all missing assignments without changing the current predictor. We evaluate this audit retrospectively in two self-triage cohorts, 108 clinicians rating three simulated psychiatric interviews, and 126 counseling demonstrations annotated by ten experts. Under shared random half histories, Fisher adaptation based on observed ratings reduces log loss relative to response frequency by 0.1071 nats in the clinician audit (95% CI [0.0867,0.1261]) and 0.0223 in counseling (95% CI[0.0121,0.0321]). The counseling test excludes target experts and source videos from training. Certification covers 5.4% and 10.5% of targets; fixed width acquisition raises coverage to 13.7% and 28.6% at the same budgets. With weights and responses held fixed, the observed-rating update also has lower loss than an affine partial update while certifying exactly the same evaluated targets. The certificate establishes agreement with the completed model, not correctness or stability at intermediate updates. These rating tasks test the method; clinical usefulness remains untested.

Abstract: Functional data analysis provides a framework for studying time-varying associations between covariates and high-frequency biological signals. In digital health, participants often contribute repeated functional trajectories, yet existing multilevel regression methods largely focus on the conditional mean, overlooking variability and its role in biological heterogeneity. We propose a heteroskedastic functional mixed-effects regression framework that jointly models the conditional mean and variance of repeated trajectories, with participant-specific random effects in both components. We also develop a participant-level bootstrap to quantify uncertainty in estimated functional parameters. We apply the method to postprandial glucose trajectories from the AEGIS study. Premeal glucose, carbohydrate intake, and meal type show distinct time-dependent associations with glucose variability, whereas HbA1c is more strongly associated with the conditional mean than with variability. Substantial between-participant differences in residual variability remain even after accounting for observed covariates. These findings show that modeling conditional variability alongside the mean provides insights into glucose regulation and individual heterogeneity that mean-focused analyses may miss. More broadly, the proposed framework enables joint characterization of average responses, variability, and participant-specific heterogeneity in repeated biological signals.

Abstract: The Human Reference Atlas (HRA) v2.5 includes 4,807 anatomical structures, 1,347 cell types, and 2,215 biomarkers (genes, proteins, lipids); 81 organs with 1,392 anatomical structures modeled in three dimensions (3D). Mapping new data into the HRA requires 3D spatial registration and/or cell type annotation (CTann). Many CTann tools exist but few use cell type names linked to Cell Ontology (CL), making comparisons across tools, tissues, and organs difficult. The HRA supports five transcriptomics CTann tools (Azimuth, CellTypist, popV, FR-Match, Pan-Human Azimuth), four sc-proteomics CTann tools (DeepCell Types, DeepCell Types-HuBMAP, Robust Image-Based Cell Annotator [RIBCA], and SpaTial cELl LeARning [STELLAR]), plus a Cell Distance Explorer (CDE) Spatial Omics collection covering previously published annotations across 12 tissue types with 47,349,496 cells. Assigned cell types differ in number, resolution, and coverage. This paper details how 10 crosswalks were constructed, validated, and used to compile a "Supertree" that can be used to compare CTann across organs and technologies in support of reference atlas constitution and usage. All data and code can be accessed at https://cns-iu.github.io/hra-ctann-supporting-information.

Abstract: Neurodegenerative disorders are characterized by their progressive nature, with distinct stages that require precise identification and tracking of progression trajectories for accurate diagnosis. However, existing diagnostic methods often fail to fully capture the unique brain patterns associated with different stages, limiting their ability to effectively target specific stages of the disorders. To address this issue, we propose a Stage-aware Multimodal Brain Graph Contrastive Learning (smBGC) method. Specifically, smBGC introduces a stage-aware feature enhancement strategy to capture stage-specific features across all progression stages, enabling the generation of stage-aware multimodal brain graph representations. In addition, smBGC incorporates a stage-adaptive brain graph augmentation approach within a graph contrastive learning framework, facilitating multimodal fusion and improving the model's capability to distinguish between different disorder stages. Experiments on two real-world datasets demonstrate the superior performance of smBGC in various neurodegenerative disorder diagnosis tasks compared with state-of-the-art methods.

Abstract: Physicians read electrocardiograms (ECGs) from waveform shape and timing, yet accurate deep-learning classifiers rarely show which waveform regions drive their predictions. We introduce LYNX-ECG, a graph neural network that represents each heartbeat as a sparse visibility graph of timestamped turning points. On CODE-Test, LYNX-ECG reaches a macro F1 score of 0.889, outperforming five baselines including the ECGFounder foundation model, while its graphs need about 14% of the storage of graphs built from every sample. Because each node maps to a specific ECG time, we can delete clinically defined waveform regions and alter beat timing to audit predictions. For conduction disorders, the clinically expected region was the strongest evidence in more than 90% of correct predictions. Bradycardia and tachycardia calls follow heart rate with conduction intervals held fixed, and atrial fibrillation calls vanish when beat-to-beat variation is removed. Sparse visibility graphs thus pair accurate ECG classification with clinically grounded, testable explanations.

Abstract: Early detection of atrial fibrillation (AFIB), characterized by irregular heart rhythms, can potentially save thousands of lives annually [20]. Numerous studies have attempted to address this challenge using various algorithms and methodologies. This paper introduces a novel and robust method that enhances the precision of AFIB detection. Neural Networks (NN) are widely recognized for their ability to identify complex electrocardiogram (ECG) patterns, facilitating the accurate detection of irregular heart rhythms. This study explores multiple conventional machine learning algorithms alongside neural network approaches to develop a robust AFIB detection system. To evaluate the effective- ness and robustness of the proposed method, the MIT BIH arrhythmia database (MITDB) and the Long-term Atrial Fibrillation Database (LTAFDB), both available on the Physionet site. Our approach presents a novel feature engineering technique, employing specific features such as a time series of the differences in successive heart rates (dBPM) rather than traditional inter-beat intervals. The model analyzes inputs labeled into five distinct classes, using only clean intervals identified through a new labeling method for training while treating others as outliers. This strategy ensures a more accurate and reliable AFIB detection process. The best-performing model, a Deep Neural Network (DNN), achieved an F1 score of 95.18% for binary atrial fibrillation classification on a benchmark dataset distinct from the training set. Models trained using our approach have demonstrated superior performance than previously published models.

Abstract: Individual patient data are the gold standard for secondary analysis of time-to-event evidence, but they are rarely released, and reconstruction from a published Kaplan-Meier figure could help. Existing methods all claim high fidelity, each judged by whichever agreements it happens to achieve, and each returns a single dataset although the map from figure to data is not bijective. We define fidelity explicitly: a dataset is admissible when it reproduces the risk table exactly, follows the digitized curve to within the image resolution, and respects the printed median, interval and event count. We propose StochIKM, a stochastic algorithm that digitizes the figure with a vision-capable language model and uses the high-fidelity definition as constraints, returning an admissible set of individual patient datasets. A solution is guaranteed to exist whenever the figure has been read correctly, so a failure indicts either the published figure or its digitization. In simulation and in 79 arms from published oncology trials, StochIKM reproduces the published risk tables and medians more closely than LLM-improved IPDfromKM, a high-impact alternative.

Abstract: As two most prevalent neurodegenerative diseases (NDDs), Alzheimer's disease (AD) and Parkinson's disease (PD) shared commonalities both in symptomology and pathophysiology. However, to what extent and in what way these two distinct disorders overlap remain underexplored. In this study, we propose an explicit, interpretable and spatially adaptive transfer learning framework to elucidate the morphometric biomarkers linked to AD and PD. We applied such high-dimensional machine learning model to two most popular open-source neuroimaging datasets, i.e., Alzheimer's Disease Neuroimaging Initiative (ADNI) and Parkinson's Progression Markers Initiative (PPMI). On the one hand, our cross-disorder classifier revealed neuro-anatomical alteration patterns well aligning with previous neuroimaging research findings and underscoring patients' clinical presentations. On the other hand, the spatially adaptive trained classifier using shared prior achieved statistically superior diagnostic performance compared to respective individual models, which further highlights the cross-disorder framework's efficacy on modeling limited-size datasets across disease domains.

Abstract: Analyzing physiological sleep signals is important for assessing sleep quality and detecting sleep-related disorders. Learning inter-modal relationships across physiological signals can provide richer representations for downstream sleep analysis. We introduce \textbf{SleepPairNet}, a multimodal self-supervised learning (SSL) framework that learns pairwise interactions between physiological signals through cross-attention. We propose a random modality-pair selection strategy that exposes the model to diverse signal pairs during pretraining, enabling flexibility across modality combinations at test time. Pretrained on the NCH dataset, SleepPairNet outperforms the evaluated SSL baselines on all respiratory tasks while remaining competitive on EEG-related tasks, and shows even stronger results when transferred to the external CHAT dataset.

Abstract: Topic models learn gene programs from single-cell RNA sequencing (scRNA-seq) data, and the same counts are then used to test which genes and donors are associated with those programs. Count splitting is the standard remedy for this reuse, but it has not been tested when an encoder reads each cell's counts. On a 104,400-cell lupus cohort with 300 permuted null genes per run, naive tests rejected 36% of null gene-topic pairs at a nominal 5% level, and the usual two-way split rejected 38%, because the encoder sees the tested gene's own held-out count. Encoding topic usage on a separate third of the counts, with beta-binomial thinning, lowered this to 7.7% (6.7% for genes outside the model); for two other models, LDVAE and scETM, the rate was 7.0-7.3% (naive 51-59%). A label-free interferon program shifts in held-out counts in the lupus cohort and in a pre-specified COVID-19 cohort, and donor-level topic usage predicts disease (AUROC 0.96) beyond cell-type proportions (0.86 and 0.58 in the two cohorts). Calibration and prediction did not depend on the model. We do not claim better clustering, prediction or topic stability. Code, figures and result tables are public (MIT license; DOI 10.5281/zenodo.23307193).

Abstract: Mixed methods research offers established strategies for integrating qualitative and quantitative evidence, but less guidance for integrating two computational representations of the same qualitative corpus. This methodological study introduces Parallel-Then-Integrated (PTI), a four-step procedure for explicit and traceable integration. We present PTI using 1,203 interview paragraphs from 29 mothers in Nepal. The two representations were ontology-guided Patient Voice (PV) coding with PVminer and data-driven themes generated through non-negative matrix factorization (NMF). A theme-complete crosswalk documented convergence, divergence, and preprocessing-sensitive differences. Depression status, based on Patient Health Questionnaire-9 (PHQ-9) scores, served as a secondary external anchor for exploratory association checks. PTI provides a structured approach for integrating computational representations of qualitative text within mixed methods research.

Abstract: Chain of Thought (CoT) monitoring is used by frontier labs to detect misbehaviour of AI systems. Yet, its reliability remains poorly understood. We attribute this to the lack of a successful organizing principle that leads to generalizable conclusions about monitoring reliability in different regimes. Recent work attributes monitoring failures to the CoT not being necessary for the answer, and argues that CoT becomes monitorable when it is necessary. We show that this binary necessary/not-necessary organizing principle is too coarse by demonstrating that monitoring failures persist even if the CoT is necessary for the answer. In response, we introduce a more fine-grained organizing principle based on a hierarchy of four CoT necessity levels. Building on this hierarchy, we develop a framework for measuring CoT monitorability at the three lowest levels via intervention experiments. We demonstrate monitoring failures at all three of these levels on realistic, hard prediction tasks. On a difficult medical-diagnosis task, patient sex and a clinical attribute influence the answer in at least 19% and 24% of correct responses with a necessary CoT, respectively, yet their influence is almost never articulated. These results show that even when CoT is necessary, it is not guaranteed to be monitorable.

Abstract: Participant-disjoint cross-validation prevents direct participant overlap, but it may still produce optimistic estimates when acquisition site and diagnosis are structurally dependent. We examined Parkinson's disease classification using 159 pressure-walkway summary features from 133 WearGait-PD participants, including 65 participants with Parkinson's disease and 68 controls. The cohort was non-factorial: two sites contained only one diagnostic group, whereas VA-Seattle was the only site containing both groups. In repeated participant-level nested cross-validation, site alone achieved an area under the receiver operating characteristic curve (AUC) of 0.955 plus or minus 0.003, exceeding the gait-feature model (0.882 plus or minus 0.014). Under the evaluated logistic-regression pipelines, adding gait features to site, age, and sex produced no measurable improvement in pooled discrimination. More site-controlled evaluations produced lower and imprecise estimates: VA-Seattle same-site AUC was 0.668 (95% bootstrap confidence interval 0.477 to 0.836), and the JHOC + Bayview-to-VA-Seattle transport stress test achieved 0.576 (0.375 to 0.770). Supporting within-diagnosis site-prediction and site-conditioned permutation analyses showed that the gait representation retained substantial acquisition-site structure. These findings indicate that participant-disjoint pooled validation did not isolate disease-related discrimination from site-associated variation. They encourage careful examination of site-label dependence and, where possible, support for pooled results through within-site or independent validation before broader claims are made about generalization across institutions.

Abstract: Randomized controlled trials (RCTs) identify treatment effects without confounding but are often small, whereas observational studies (OBS) are large but may be confounded. Many estimators combining a small RCT with a large OBS have been developed for the average treatment effect (ATE) and the conditional ATE (CATE). However, existing ATE estimators either make assumptions on the OBS or do not borrow enough power from them. The CATE has been studied less than the ATE. Existing CATE methods either assume the OBS are unconfounded, rely on a model of the confounding function, or accept bias in exchange for lower variance. We therefore propose a framework that, without special assumptions on the OBS, fuses the OBS and the RCT by preserving the unbiasedness of RCT-based estimation while borrowing power from the large OBS to boost precision. Applying this principle, we build an ATE estimator, AIPW-Fusion, with closed-form weights and confidence intervals, and two CATE learners, DR-Fusion and R-Fusion. Experiments corroborate our findings.

Abstract: Multimodal depression assessment offers a promising approach to analyzing behavioral patterns associated with depression. However, existing methods often rely on dense and opaque multimodal representations, making it difficult to interpret the behavioral patterns underlying their predictions. In this work, we introduce BehavDep, a sparse factor-based framework that decomposes multimodal behavioral representations into sparse latent factors and associates them with behaviorally meaningful concepts through a semantic bridge. To address the mismatch between user-level annotations and heterogeneous video-level behaviors, BehavDep further learns video-level depression tendency scores under weak supervision and aggregates information across multiple observations for user-level assessment. Extensive experiments demonstrate that BehavDep achieves the best overall assessment performance while revealing complementary modality contributions, heterogeneous behavioral patterns across observations, and prediction responses to concept-level editing. These results show that BehavDep provides a structured and interpretable approach to analyzing multimodal behavioral representations for depression assessment.

Abstract: Depression has no diagnostic blood test. Language models promise tireless, consistent assessment, but can accurate raters disagree about individuals? We pre-registered 880 language-model raters, crossing 11 open models with prompting and scoring choices, and applied them to 189 interviews against the eight-item Patient Health Questionnaire. Model choice explained 30.0% of summed-symptom score variance, stable participant differences 10.5%. Two randomly drawn raters with area under the receiver operating characteristic curve (AUC) >= 0.70 disagreed on screening decisions for 40% of participants, on average. Average over-rating governed how many were flagged, yet equal-capacity raters chose differently for about one participant in five. A locked analysis of 86 new interviews reproduced the main pre-registered findings. Exploratory recalibration with 40 labelled participants raised accuracy from about 60% to 75% and halved disagreement, leaving one participant in five decided differently. Calibration repaired much of the rater dependence without securing agreement about individuals.

Abstract: Natural language processing (NLP) models can detect depression-related language in text written near the time symptoms are measured, but whether pretrained transformers can predict depressive symptoms from text written twelve years earlier is largely untested. In the National Child Development Study, a British birth cohort, we predict probable depressive symptoms at age 23 from essays the same people wrote at age 11. Our baseline, a logistic regression on six childhood covariates, outperforms every text model that sees only the essay: seven fine-tuned transformers, a bag-of-words model, frozen embeddings and four zero-shot large language models. Its area under the receiver operating characteristic curve (AUC-ROC) is 0.737 against 0.670 for the best transformer on the primary seed, and no added text score detectably raises the baseline's AUC-ROC. None of the five domain-pretrained transformers detectably beats its general-domain control after Bonferroni correction. For long-horizon prediction, the baseline remains the model to beat.

Abstract: The high cost of functional molecular assays, and prevalence of missing modalities and unmatched samples in computational biology, create significant barriers to comprehensive multi-omic profiling, essential for capturing and reasoning over molecules, cells, tissues, and organisms. This work proposes a model that learns meaningful representations from multi-omics cancer data supporting the reconstruction of missing and unpaired modalities. Contrary to increasingly complex, larger models, e.g. Foundation Models (FMs), ARO prioritizes practical applicability in limited or incomplete data settings. ARO optimally reconstructs missing modalities (MSE of $0.15$ on the validation and test data in the Unmasked settings), with its learned latent embeddings enabling a downstream cancer classification task. Our findings indicate that analyzing diverse molecular layers as a single integrated system offers a reliable and cost-efficient approach, reducing dependence on large-scale experimental testing, while still supporting multi-omic exploration in limited data settings.

Abstract: Exploring high-order correlations and long-range dependencies in brain networks holds significant value for both neuroscience research and clinical diagnosis. However, previous studies have lacked a unified integration of high-order and long-range dependency information in brain networks, and there is substantial redundancy behind various types of information. These issues limit their effectiveness in the diagnosis of brain diseases. To address this, we propose an Information Bottleneck-Guided Adaptive HyperGraph Transformer (IBAHGT). By incorporating the information bottleneck (IB) principle, this approach enables adaptive learning of high-order correlations and both short- and long-range dependencies within a unified framework for brain network analysis, achieving high-precision brain disease diagnosis. IBAHGT consists of three key components: an information bottleneck-guided adaptive hypergraph convolution, which introduces a novel hypergraph information bottleneck (HIB) principle to adaptively learn hypergraph message-passing weights between nodes and hyperedges, optimizes information flow and captures high-order information in brain networks that is maximally informative and minimally redundant (MIMR). The Transformer encoder captures global information within brain networks through the attention mechanism, specifically modeling short- and long-range dependencies. An information bottleneck-guided node-level adaptive fusion employs the IB principle to learn independent weights for each node, facilitating the fine-grained integration of high-order information and global information to obtain an efficient representation for downstream tasks. Extensive experiments demonstrate that the proposed method outperforms current state-of-the-art methods and can identify biomarkers for clinical applications.

Abstract: Understanding how individual brain development unfolds over childhood requires modeling developmental trajectories from sparse longitudinal observations. Long-term neurodevelopmental forecasting is challenging because each child is typically observed at only a few irregularly spaced visits, while developmental dynamics vary across individuals and age. Generic continuous-time models accommodate irregular timing but often absorb these factors into a single flexible transition function, providing little structure for how population progression, individual variability, and developmental age shape the dynamics. We propose DevelopmentODE, a structured continuous-time framework that organizes population- and subject-specific variation within a shared developmental geometry while allowing the governing dynamics to evolve with age. The model builds this geometry around a developmental canal representing the population trajectory, whose local direction provides a reference for organizing subject-specific variation. Subject deviation velocities are constrained relative to this direction, while a shared nonlinear deviation field captures individual developmental motion without disrupting population-level progression. DevelopmentODE further models developmental non-stationarity through ordered age-dependent deformations of the shared vector field, progressively adapting a common dynamical structure as age changes, while elapsed time determines the integration horizon. This formulation uses population-level developmental structure to guide learning from sparse individual trajectories while allowing dynamics to evolve smoothly with age. We evaluate DevelopmentODE on longitudinal fMRI by predicting future functional connectivity of the same child from earlier observations. DevelopmentODE consistently outperforms competing baselines across short- and long-horizon predictions.

Abstract: Electroencephalography (EEG)-based closed-loop neuromodulation calls for a subject-specific structured state, as opposed to a single disease probability, specifying which brain regions are deviant, at which frequencies, and at which lags. Sensor-space models keep the strongest diagnostic evidence without anatomy, source-space models give anatomy at a loss of predictive signal, and post-hoc attributions stay outside the prediction. We separate the two instead of forcing them into one representation, and propose DMD-EEG (Dual-view Multiscale Deformation for EEG), which keeps a fixed scalp spectral expert for diagnosis and models the source-space disease-related representation as a low-rank, sparse, iterative deformation of a healthy neural-dynamics prior in a $46$-region-of-interest (ROI) $\times$ $5$-frequency $\times$ $4$-lag (autocorrelation-timescale) space. The two experts meet only at a fixed decision level, so the source state is architecturally separate from the scalp expert. Across major depressive disorder (MDD), first-episode psychosis (FEP), and Parkinson's disease (PD), decision-level fusion matches the strongest single expert on MDD and FEP and exceeds the source branch on PD. On FEP the source expert is the strongest branch, the task where the deformation contributes most. The source state is an explicit ROI-frequency-lag attribution defined in a shared source coordinate system across montages, which we treat as an anatomically-coordinated predictive representation whose coordinates are directly readable and hypothesis-generating. The highest-saliency coordinates align with established disease circuitry (fronto-limbic-temporal regions in MDD, motor-cortex beta in PD), and the MDD state transfers by rank to an unseen cohort recorded with a different montage.

Abstract: Impaired glucose regulation is a major contributor to metabolic dysfunction and type 2 diabetes. This study developed an interpretable machine-learning framework to predict log-transformed fasting blood glucose using metabolic, hormonal, lifestyle, demographic, nutritional, and circadian variables from the National Health and Nutrition Examination Survey 2017--2020 pre-pandemic dataset. After merging multiple NHANES sub-datasets, data processing used a leakage-resistant pipeline in which imputation, scaling, and one-hot encoding were performed only after dataset splitting and within training folds. Elastic Net, LASSO, and XGBoost models were evaluated using 94 candidate predictors and engineered circadian interaction terms. Performance was assessed using mean absolute error, root mean squared error, coefficient of determination, calibration, and Shapley Additive Explanations. The final interaction-augmented XGBoost model achieved strong performance on the independent test set, with a mean absolute error of 0.0804, a root mean squared error of 0.1148, and a coefficient of determination of 0.7761, using 10 predictors. Glycohemoglobin was the dominant predictor, followed by insulin, diabetes diagnosis, gamma-glutamyl transferase, age, race, and gender. Among the engineered interaction terms, sleep midpoint multiplied by age was consistently retained in repeated random-split analyses, although its contribution remained modest relative to dominant glycaemic predictors. These findings support further investigation of circadian-age interactions in metabolic health.

Abstract: Accurately isolating disease-related features from confounding covariates (e.g., age, gender, site) and individual variations remains a fundamental challenge in medical image classification. Traditional regression-based approaches may ignore non-linear relations between image features and true covariates. To overcome this issue, we present a generalized Medical Imaging Disentanglement Learning (MedIDL) framework. MedIDL maps image features into three mutually orthogonal latent spaces through specialized disentanglement heads: a disease classification head guided by a supervised loss, a covariate-alignment head constrained by cross-subject similarity matching, and a Gaussian head absorbing individual variations. We evaluated our framework across 7 datasets encompassing diverse imaging modalities. MedIDL outperforms state-of-the-art supervised and self-supervised classification methods in accuracy across all datasets. Association analyses demonstrate that MedIDL successfully isolates target-specific latent representations. Gradient-based interpretability mappings localize pathognomonic patterns aligning with established clinical literature.

Abstract: Automated biomedical evidence synthesis depends on retrieving published studies, but the biomedical literature is systematically skewed toward positive findings. Deeper retrieval can therefore make a system \emph{more} likely to falsely infer benefit when the true effect is null. We formalise this phenomenon as \emph{evidence drift} and prove that, under a standard publication-bias model, the false-positive probability on null-effect queries follows a strictly increasing large-sample envelope in retrieval depth, approaching one. Empirically, on a held-out test set of 140 Cochrane-derived queries, drift rises monotonically from 7.9\% to 15.7\% as the retrieval budget grows from 3 to 20 steps, and concentrates in the null-effect class. We present DACG-agent, a drift-aware causal-graph agent that incrementally builds a causal knowledge graph from PubMed abstracts and applies a two-layer stopping policy with complementary roles: a KL-divergence monitor that detects posterior convergence (the accuracy layer), and a Bradley--Terry process reward model (PRM) whose online decline detection halts retrieval once evidence quality peaks (the efficiency layer). Against full-budget retrieval, DACG-agent reduces evidence drift from 15.7\% to 6.4\% and improves null-effect accuracy by 21 percentage points (40.0\%$\to$61.4\%) while using 67\% fewer retrieval steps; overall accuracy rises from 61.4\% to 69.3\% (95\% CI 61--77). A simulation confirms the drift result transfers from the analysed vote-counting aggregator to the deployed noisy-OR one.