Lopez-Carbonero, J. I. | Search bioRxiv

bioRxiv2026

Compact longitudinal representations derived from mixed-format lifestyle questionnaires outperform static text-derived features for ALS-versus-control classification

BackgroundMixed-format lifestyle questionnaires contain both structured variables and free-text responses, but it remains unclear whether language-derived variables provide incremental predictive value beyond structured data, and under which representational condition. It was investigated whether variables derived from patient-reported free text improve ALS-versus-control classification beyond structured questionnaire data, and whether their value depends on how temporal information is represented. MethodsA leakage-free machine-learning pipeline was developed to classify ALS versus controls from questionnaire-derived data, including a schema-guided LLM-based text-to-table extraction and a compact longitudinal encoding strategy. Three feature configurations were compared: Pool1, containing structured baseline variables only; Pool2, adding compact summaries derived from first-time-point (T1) free-text responses; and Pool3, further incorporating compact descriptors of change between T1 and T2. Logistic Regression, linear Support Vector Classification, and Random Forest were evaluated using repeated stratified holdout (10 seeds) and repeated stratified 5-fold cross-validation. Final ablation analyses were performed to isolate the contribution of the compact text block and the compact temporal block. ResultsAfter leakage correction, performance estimates became more conservative, indicating that previous results had been optimistic. In the final configuration, Pool3 achieved the best performance, with Random Forest reaching a holdout accuracy of 0.673, F1-weighted score of 0.666, and Matthews correlation coefficient of 0.323; cross-validated F1-weighted score and Matthews correlation coefficient were 0.654 and 0.312, respectively. Pool2 did not show a robust improvement over Pool1. Ablation analysis showed that removing the compact temporal block markedly reduced Pool3 performance, whereas removing the compact text block had little overall effect. These findings indicate that the primary value of language-based processing in small clinical cohorts lies not in static feature enrichment, but in enabling compact representations of longitudinal change. ConclusionsIn this setting, the main predictive gain did not arise from static text-derived variables alone, but from representing questionnaire information as compact longitudinal change descriptors. These findings suggest that, in small clinical cohorts, the value of language-based processing may lie more in summarizing trajectories than in expanding static feature spaces.

bioinformatics↗

bioRxiv2023

TDP-43-M323K causes abnormal brain development and progressive cognitive and motor deficits associated with mislocalised and increased levels of TDP-43.

TDP-43 pathology is found in several neurodegenerative disorders, collectively referred to as "TDP-43 proteinopathies". Aggregates of TDP-43 are present in the brains and spinal cords of >97% of amyotrophic lateral sclerosis (ALS), and in brains of [~]50% of frontotemporal dementia (FTD) patients. While mutations in the TDP-43 gene (TARDBP) are usually associated with ALS, many clinical reports have linked these mutations to cognitive impairments and/or FTD, but also to other neurodegenerative disorders including Parkinsonism (PD) or progressive supranuclear palsy (PSP). TDP-43 is a ubiquitously expressed, highly conserved RNA-binding protein that is involved in many cellular processes, mainly RNA metabolism. To investigate systemic pathological mechanisms in TDP-43 proteinopathies, aiming to capture the pleiotropic effects of TDP-43 mutations, we have further characterised a mouse model carrying a point mutation (M323K) within the endogenous Tardbp gene. Homozygous mutant mice developed cognitive and behavioural deficits as early as 3 months of age. This was coupled with significant brain structural abnormalities, mainly in the cortex, hippocampus, and white matter fibres, together with progressive cortical interneuron degeneration and neuroinflammation. At the motor level, progressive phenotypes appeared around 6 months of age. Thus, cognitive phenotypes appeared to be of a developmental origin with a mild associated progressive neurodegeneration, while the motor and neuromuscular phenotypes seemed neurodegenerative, underlined by a progressive loss of upper and lower motor neurons as well as distal denervation. This is accompanied by progressive elevated TDP-43 protein and mRNA levels in cortex and spinal cord of homozygous mutant mice from 3 months of age, together with increased cytoplasmic TDP-43 mislocalisation in cortex, hippocampus, hypothalamus, and spinal cord at 12 months of age. In conclusion, we find that Tardbp M323K homozygous mutant mice model many aspects of human TDP-43 proteinopathies, evidencing a dual role for TDP-43 in brain morphogenesis as well as in the maintenance of the motor system, making them an ideal in vivo model system to study the complex biology of TDP-43.

neuroscience↗