Journal: BMC Cancer

Article Title: A multi-branch ensemble learning framework for detection of non-small cell lung cancer via T-cell receptor sequencing

doi: 10.1186/s12885-026-15714-y

Figure Lengend Snippet: External test performance on DB2 (Adaptive Biotechnologies immunoSEQ platform, N = 45, Mean AUC = 0.941 ± 0.021). A ROC curves for 5-fold external test on DB2, where the training set ( N = 235) was split into 5 folds for training/validation while DB2 remained fixed as the external test set. Individual fold AUCs range from 0.913 to 0.973, with the mean ROC curve (blue, AUC = 0.941 ± 0.021) demonstrating consistent cross-platform generalization. B AUC values by fold with the mean AUC indicated by the dashed red line. DB2 consists of Sun Yat-sen new cohort (30 NSCLC) and Healthy_Cohort_2 (15 controls), using Adaptive Biotechnologies immunoSEQ platform

Article Snippet: To evaluate multi-center generalizability, the model was tested on two independent external cohorts: DB1 (Illumina MiSeq, N = 47) and DB2 (Adaptive Biotechnologies immunoSEQ, N = 45).

Techniques: Biomarker Discovery

Journal: BMC Cancer

Article Title: A multi-branch ensemble learning framework for detection of non-small cell lung cancer via T-cell receptor sequencing

doi: 10.1186/s12885-026-15714-y

Figure Lengend Snippet: External test performance on DB2 (Adaptive Biotechnologies immunoSEQ platform, N = 45, Mean AUC = 0.941 ± 0.021). A ROC curves for 5-fold external test on DB2, where the training set ( N = 235) was split into 5 folds for training/validation while DB2 remained fixed as the external test set. Individual fold AUCs range from 0.913 to 0.973, with the mean ROC curve (blue, AUC = 0.941 ± 0.021) demonstrating consistent cross-platform generalization. B AUC values by fold with the mean AUC indicated by the dashed red line. DB2 consists of Sun Yat-sen new cohort (30 NSCLC) and Healthy_Cohort_2 (15 controls), using Adaptive Biotechnologies immunoSEQ platform

Article Snippet: Fig. 5 External test performance on DB2 (Adaptive Biotechnologies immunoSEQ platform, N = 45, Mean AUC = 0.941 ± 0.021).

Techniques: Biomarker Discovery

Journal: Medecine sciences : M/S

Article Title: [Nobel Prize in chemistry 2024: David Baker, Demis Hassabis et John M. Jumper. The revolution of artificial intelligence in structural biology].

doi: 10.1051/medsci/2025060

Figure Lengend Snippet: Figure 2. Structure homodimérique de l’intégrase (IN) du virus VIH-1 (en haut à gauche) prédite par AlphaFold-3 à partir de sa séquence et d’ions liés (panneau bleu en bas). La protéine est représentée en mode ruban et est colorée en fonction de son score de confiance local plDDT. Cette protéine modulaire de 288 résidus est constituée de trois domaines : un domaine amino-terminal (NTD, 1-47) un domaine central catalytique (CCD, 59-202) et un domaine carboxy-terminal (CTD, 220-270) [39]. Le NTD lie un ion Zn2+ et le CCD deux ions Mg2+. Les structures des fragments NTD-CCD et CCD-CTD ont été résolues expérimentalement sous forme d’homodimère pour l’IN seule. La structure prédite de l’IN homodimérique entière montre de bonnes statistiques globales avec un pTM (predicted template modeling score) de 0.76 (>0.5 attendu) et un ipTM (interface pTM-score) de 0.75 (>0.6 attendu). L’analyse du PAE (predicted aligned error) (en haut à droite) montre une bonne rigidité au sein de chaque domaine (zones vert sombre) ainsi qu’entre les domaines CCD des deux monomères. La position du CTD par rapport au CCD et au NTD est légèrement flexible (zone vert clair) au sein de chaque monomère. La boucle reliant le NTD au CCD et la queue carboxy-terminale (271-288) sont plus flexibles (zones quasi-blanche). Ceci est confirmé par les couleurs jaune et orange de ces deux segments sur la représentation de la protéine en mode ruban (score de confiance plDTT localement bas à très bas).

Article Snippet: Dans la foulée, une base de données des structures prédites nommée AlphaFold DB2 a été créée par Google DeepMind et l’EMBL-EBI (EMBL’s european bioinformatics institute), afin de couvrir l’ensemble des protéines de séquence connue et ainsi accélérer la recherche.

Techniques: Virus

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Summary of datasets utilized in the study, including their key characteristics and specifications.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques: Sampling

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Comparison of gesture classification results based on sEMG data across different methods.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques: Comparison

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Classification performance of hand movement and gesture recognition models under varying noise levels in EMG signals.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques:

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Sample confusion matrices generated from the analysis of datasets using the proposed approach: ( A ) BioPatRec DB1, ( B ) BioPatRec DB3, and ( C ) Mendeley.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques: Generated

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Normalized confusion matrices showing two random results obtained from applying the proposed algorithm to BioPatRec DB2 data.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques:

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Normalized confusion matrix showcasing the classification performance of the proposed structure on the Ninapro DB2 dataset.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques:

Journal: Scientific Reports

Article Title: Hand gestures classification of sEMG signals based on BiLSTM-metaheuristic optimization and hybrid U-Net-MobileNetV2 encoder architecture

doi: 10.1038/s41598-024-82676-1

Figure Lengend Snippet: Comparison of classification model accuracy rates, presented as percentage values. Bold values indicate the highest accuracy achieved for each dataset.

Article Snippet: Compared to other models that used only the lowest number of parameters, our model achieved competitive classification accuracy for Mendeley, BioPatRec DB2, NinaPro DB5, and NinaPro DB4.

Techniques: Comparison