Biology subjects

Alam, M. N.

Publications and source records attributed to Alam, M. N..

8 recordsLinked to original sources

bioRxiv2026

RadMultiBench: A Reproducible Dual-Task Benchmark Exposing the Limits of Pre-Trained Vision-Language Models Across Thoracic and Breast Imaging

We present RadMultiBench, a multimodal benchmark of 6,148 medical images paired with expert radiological impressions across five categories--Breast Cancer, Lung Benign, Lung Malignant, Pneumonia, and Tuberculosis--targeting two core clinical tasks: automated report generation and disease classification. To enable reliable, repeatable experimentation, we provide a reproducible pipeline for data ingestion, de-identification, split management, and standardized reporting of text and classification metrics, facilitating dataset versioning and continuous benchmark testing. We compare three architectures: (1) a classical CNN-RNN with Bahdanau attention (ResNet-LSTM), (2) a general-domain vision-language (V-L) prefix-projection model (CLIP-GPT), and (3) a domain-specific V-L prefix-projection model (BioViL-T). Results reveal a sharp performance divide: the ResNet-LSTM attains near-perfect classification (99.5% accuracy) and is the only model to generate coherent reports (BLEU-4 = 0.021), whereas both prefix-projection models catastrophically fail generation (BLEU-4 [≤] 0.0001); moreover, the domain-specific BioViL-T also fails classification (11.5% accuracy). From a data-engineering angle, the pipeline is productionizable with lineage, feature registries, and continuous integration to guard against silent regressions when updating models or data; from a data-science angle, the benchmark provides baselines and analyses that foreground feature alignment and architectural suitability as first-order concerns, inviting ablations on multimodal objectives, retrieval-augmented encoders, and uncertainty-aware decoding. Overall, RadMultiBench highlights that state-of-the-art status in general V-L domains does not translate automatically to high-stakes clinical tasks, and it furnishes the tooling and evidence base needed to develop robust, clinically viable multimodal systems.

bioinformatics↗

bioRxiv2026

Participatory Systems Mapping and Experimental Games to Explore Biosecurity Adoption in Broiler Production in Bangladesh

IntroductionAntimicrobial Resistance (AMR) presents a critical public health challenge, particularly in smallholder broiler farming, where antibiotics are often used preventively in the absence of effective biosecurity measures. ObjectiveThis study investigates the adoption of biosecurity practices as a sustainable alternative to antibiotics through Participatory Systems Mapping and Experimental Games. MethodsA participatory mixed-methods study was conducted in southern Bangladesh (September 2024-June 2025). Causal Loop Diagrams (CLDs) were co-created with farmers, dealers, and veterinary officers. Ten broiler farmers from single village were selected via purposive and snowball sampling. Experimental games simulated four production cycles where farmers chose Option A (biosecurity, adopters) or Option B (antibiotics, non-adopters) after several interactive trainings. Key metrics including biosecurity compliance (0-12 scale), mortality, FCR, antibiotic use, outbreak history, and economic outcomes were recorded. ResultsCLD analysis revealed a reinforcing loop of increased antibiotic reliance driven by fear of mortality, and balancing loops involving training, biosecurity practices, and consumer incentives to reduce use. Five farmers chose Option A, and both groups remained stable until Round 4. Adopters had flock sizes of 800-2000 birds (non-adopters, 600-1000; mean for both = 1000), were younger, and more educated compared to non-adopters. At baseline, both groups had similar biosecurity scores (0). Adopters had higher mean outbreaks (2 vs. 1.4), mortality (5.6 vs. 4.2), antibiotic use (3.6 vs. 3), and FCR (1.8 vs. 1.6) compared to non-adopters. By Round 4, adopters improved biosecurity scores by 125%, eliminated outbreaks, reduced mortality by 52.6%, stopped antibiotic use, improved FCR by 13.3%, and gained 71.72% profit per bird compared to non-adopters. Non-adopters, influenced by adopters, increased biosecurity scores by 25%, reducing outbreaks, mortality, antibiotic use, and FCR. Adopters also increased direct sales to consumers, yielding a 10%-16% profit gain per bird each round. ConclusionThis study highlights the successful adoption of biosecurity practices by farmers, replacing antibiotics and improving production outcomes. Farmer-driven adoption of these practices fosters long-term sustainability and supports a healthier planet within the One Health framework.

microbiology↗

bioRxiv2026

Integrated physiological performance and Nax1-mediated sodium exclusion reveal mechanisms of salinity tolerance in spring wheat (Triticum aestivum L.)

Soil salinity is a rapidly intensifying abiotic stress that significantly limits wheat productivity, particularly in coastal and irrigated agroecosystems. Although sodium (Na+) ion exclusion has been recognized as a key tolerance mechanism, the integration of physiological performance with Nax1-mediated molecular regulation among regionally adapted wheat genotypes remains insufficiently characterized. The present study aimed to dissect salinity tolerance by combining hydroponic phenotyping, multivariate trait analysis, molecular marker profiling, and quantitative expression analysis of the Na+ ion transporter gene Nax1. Seventeen spring wheat genotypes were evaluated under four salinity levels (0.0, 10, 12, and 14 dS m-{superscript 1}). Germination and survival rate, shoot and root growth, and biomass accumulation were measured. Principal component analysis (PCA) and hierarchical clustering were performed to classify genotypes, while SSR (simple sequence repeat) and Nax-linked markers assessed genetic diversity. Relative Nax1 expression was quantified using qRT-PCR (quantitative real-time polymerase chain reaction). Salinity significantly reduced germination, survival, elongation, and biomass, with strong genotype-dependent variation. Multivariate analyses clearly separated tolerant and sensitive genotypes, with biomass retention and survival contributing most to total variation. Marker analysis revealed moderate genetic polymorphism. Notably, tolerant genotypes exhibited 3-6-fold induction of Nax1 under severe salinity, positively correlating with biomass maintenance. These findings demonstrate that salinity tolerance in wheat is associated with coordinated physiological resilience and enhanced Nax1-mediated Na ion exclusion, thereby advancing mechanistic understanding and supporting molecular-assisted breeding for salt-affected environments.

plant biology↗

bioRxiv2025

Compact Vision--Language Models Enable Efficient and Interpretable Automated OCT Analysis Through Layer Specific Multimodal Learning

Translating the intricate anatomical signatures of retinal disease from OCT B-scans into clear, accurate clinical narratives demands AI models that seamlessly fuse visual features with domain expertise. We curated a multimodal dataset of 40,000 OCT B-scans from public repositories and private clinical cohorts, each paired with expert validated summaries spanning six conditions: diabetic macular edema, diabetic retinopathy, geographic atrophy, drusen, choroidal neovascularization, and healthy retina. We introduce LO-VLM, a compact (247M parameter) vision-language model (VLM) that infuses anatomical guidance into both encoder and decoder for free form summary generation and multiclass disease classification. Benchmarking against state-of-the-art RetinaVLM, LLaVA-Med, and a ViT vision only model demonstrates superior performance. In a blinded evaluation by three board certified retina specialists scored the generated summaries, LO-VLM narratives achieved mean = 8.5 (standard deviation = 1.15) out of 10, compared to mean = 5.5 (standard deviation = 1.13) for RetinaVLM (p < 0.0001). In quantitative evaluations, LO-VLM achieved an SBERT similarity of 0.803 and a BERTScore F1 of 0.715, representing improvements of 8.2% and 28.8% over specialized VLM baselines. For disease classification, LO-VLM reached 96% accuracy (F1 = 96%), outperforming ViT by 13% and exceeding medical VLM benchmarks by over 62%. By reconciling interpretability with computational efficiency, LO-VLM establishes a new paradigm for efficient AI models in OCT interpretation.

bioengineering↗

bioRxiv2024

EYE-Llama, an in-domain large language model for ophthalmology

1BackgroundTraining Large Language Models (LLMs) with in-domain data can significantly enhance their performance, leading to more accurate and reliable question-answering (Q&A) systems essential for supporting clinical decision-making and educating patients. MethodsThis study introduces ophthalmic LLMs trained on in-domain, well-curated datasets. We present an open-source substantial ophthalmic language dataset for model training. Our models (EYE-Llama), were pre-trained on an ophthalmology-specific dataset, including paper abstracts, textbooks, and Wikipedia articles. Subsequently, the models underwent fine-tuning using a diverse range of QA pairs. Our models were compared to baseline Llama 2, ChatDoctor, Meditron, Llama 3, and ChatGPT (GPT3{middle dot}5) models, using four distinct test sets, and evaluated quantitatively (Accuracy, F1 score, BERTScore, BARTScore and BLEU score) and qualitatively by two ophthalmologists. FindingsUpon evaluating the models using the synthetic dialogue test set with three different metrics (BERTScore, BARTScore, and BLEU score), our models demonstrated superior performance. Specifically, when evaluated using BERTScore, our models surpassed Llama 2, Llama 3, Meditron, and ChatDoctor in terms of F1 score, and performed on par with ChatGPT, which was trained with 175 billion parameters (EYE-Llama: 0.57, Llama 2: 0.56, Llama 3: 0.55, Meditron: 0.50, ChatDoctor: 0.56, and ChatGPT: 0.57). Additionally, the EYE-Llama model outperformed the above models when evaluated using BARTScore and BLEU scores. When tested on the MedMCQA test set, the fine-tuned models exhibited higher accuracy compared to Llama 2, Meditron, and ChatDoctor models (EYE-Llama: 0.39, Llama 2: 0.33, ChatDoctor: 0.29, Meditron: 0.22). However, ChatGPT, and Llama 3 models outperformed EYE-Llama, achieving accuracies of 0.55, 0.78, and 0.90, respectively. On the PubmedQA test set, our model showed improved accuracy over all other models (EYE-Llama: 0.96, Llama 2: 0.90, Llama 3: 0.92, Meditron: 0.76, ChatGPT: 0.93, ChatDoctor: 0.92). InterpretationThe study shows that pre-training and fine-tuning LLMs like EYE-Llama enhances their performance in specific medical domains. Our EYE-Llama models surpass baseline Llama 2 in all evaluations, highlighting the effectiveness of specialized LLMs in medical QA systems. FundingFunded by NEI R15EY035804 (MNA), R21EY035271 (MNA), and UNC Charlotte Faculty Research Grant (MNA)

bioengineering↗

bioRxiv2023

Federated learning for diagnosis of age-related macular degeneration

This paper presents a federated learning (FL) approach to train deep learning models for classifying age-related macular degeneration (AMD) using optical coherence tomography image data. We employ the use of residual networks and vision transformer encoders for the normal vs AMD binary classification, integrating four unique domain adaptation techniques to address domain shift issues caused by heterogeneous data distribution in different institutions. Experimental results indicate that FL strategies can achieve competitive performance similar to centralized models even though each local model has access to a portion of the training data. Notably, Adaptive Personalization FL strategy stood out in our FL evaluations, consistently delivering high performance across all tests due to its additional local model. Furthermore, the study provides valuable insights into the efficacy of simpler architectures in image classification tasks, particularly in scenarios where data privacy and decentralization are critical using both encoders. It suggests future exploration into deeper models and other FL strategies for a more nuanced understanding of these models performance.

bioengineering↗

bioRxiv2020

Evaluation of Neuropharmacological Effects of Different Chemical Extracts of Flemingia Stricta (Roxb.) Leaves.

BackgroundTraditional preparation of the leaf of Flemingia stricta (Fabaceae) Roxb., a medicinal plant of the Indian subcontinent, has been used for treatment of different diseases as herbal preparation. Our purpose was to analyze Neuropharmacological effects of different chemical extracts of Flemingia stricta Roxb. as particular form of behavioral inhibition that occurs in response to novel environmental events. MethodsIn present study, the anxiogenic activity of crude extracts of Flemingia stricta leaves was determined using standard animal behavioral models, such as hole cross and open field; Sedative property and anxiolytic potential were assessed by conducting thiopental sodium induced sleeping times tests and elevated plus maze test respectively. ResultsThe crude extracts at both dose (200 and 400mg/kg) exhibited a significant (P<0.05, P<0.01) dose-dependent suppression of motor activity and exploratory activity of mice in both open field and hole cross test. In anxiolytic study, extracts displayed increased percentage of entry into open arm at the dose of 200 and 400mg/kg. Extracts produced a significant (P<0.05, P<0.01) increase in sleeping duration and reduction of onset of sleep compared to sodium thiopental at both doses (200 and 400mg/kg). ConclusionThis study demonstrates that the treated extracts have significant central nervous system depressant effect. Further studies on the active constituent of the extract can provide approaches for therapeutic intervention.

pharmacology and toxicology↗

bioRxiv2020

Antimicrobial Resistance Profile and mcr-1 gene Detection in Salmonella Isolates from Poultry in Bangladesh: Molecular and Bioinformatics Characterization

Antimicrobial resistance gene mcr-1 has been disseminated globally since its first discovery in Southern China in late 2015. However, the mcr-1 gene had not been identified previously in Salmonella isolates from poultry in Bangladesh. Here, we aimed to explore antimicrobial resistance gene mcr-1 in Salmonella isolates. Eighty two Salmonella isolates were isolated and characterized from suspected poultry specimens received from different zones of the country. A phenotypic disc diffusion assay with 15 antimicrobial agents was performed following CLSI standard. The disk diffusion assay showed that, all of the isolates presented high resistance to colistin (92.68%), oxytetracycline (86.59%), co-trimoxazole (76.83%), ciprofloxacin (73.17%) and enrofloxacin (65.85%). Further, randomly selected 10 Salmonella isolates were analyzed by polymerase chain reaction (PCR) targeting genus-specific invA and antimicrobial (colistin) resistance mcr-1 genes. Five were confirmed for the presence of the mcr-1 gene belonging to Salmonella spp. Further, sequencing followed by phylogenetic analysis revealed divergent evolutionary relation between the LptA and MCR proteins rendering them resistant to colistin. Three-dimensional homology structures of MCR-1 proteins were constructed and verified using different bioinformatics tools. Moreover, molecular docking interactions suggested that, MCR-1 and LptA share a similar substrate binding cavity which could be validated for the functional analysis. The results represent here is the first molecular and in silico analysis of colistin resistance mcr-1 gene of Salmonella in poultry in Bangladesh, which may emphasize the importance of the study on antibiotic resistance genes requiring for national monitoring and strategic surveillance in the country.

microbiology↗