Genome-scale exon perturbation screens uncover exons critical for cell fitness

Molecular Cell, Journal Year: 2024, Volume and Issue: 84(13), P. 2553 - 2572.e19

Published: June 24, 2024

CRISPR-Cas technology has transformed functional genomics, yet understanding of how individual exons differentially shape cellular phenotypes remains limited. Here, we optimized and conducted massively parallel exon deletion splice-site mutation screens in human cell lines to identify that regulate fitness. Fitness-promoting are prevalent essential highly expressed genes commonly overlap with protein domains interaction interfaces. Conversely, fitness-suppressing enriched nonessential genes, exhibiting lower inclusion levels, intrinsically disordered regions disease-associated mutations. In-depth mechanistic investigation the screen-hit TAF5 alternative exon-8 revealed its is required for assembly TFIID general transcription initiation complex, thereby regulating global gene expression output. Collectively, our orthogonal perturbation established a comprehensive repository phenotypically important uncovered regulatory mechanisms governing fitness expression.

Language: Английский

---

Automatic sleep stage classification using deep learning: signals, data representation, and neural networks

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 57(11)

Published: Sept. 23, 2024

Language: Английский

---

Neural dynamics of shifting attention between perception and working-memory contents

Proceedings of the National Academy of Sciences, Journal Year: 2024, Volume and Issue: 121(47)

Published: Nov. 13, 2024

In everyday tasks, our focus of attention shifts seamlessly between contents in the sensory environment and internal memory representations. Yet, research has mainly considered external isolation. We used magnetoencephalography to compare neural dynamics shifting visual within vs. domains. Participants performed a combined perception working-memory task which two sequential cues guided upcoming (external) or memorized (internal) information. Critically, second cue could redirect content same alternative domain as first cue. Multivariate decoding unveiled distinct patterns human brain activity when Brain distinguishing within- from between-domain was broadly distributed highly dynamic. Intriguingly, crossing domains did not invoke an additional stage prior attention. Alpha lateralization, canonical marker spatial attention, showed no delay redirected domain. Instead, evidence suggested that states associated with given linger influence subsequent Our findings provide critical insights into govern attentional working memory.

Language: Английский

---

Application of Machine Learning Models in Social Sciences: Managing Nonlinear Relationships

Encyclopedia, Journal Year: 2024, Volume and Issue: 4(4), P. 1790 - 1805

Published: Nov. 27, 2024

The increasing complexity of social science data and phenomena necessitates using advanced analytical techniques to capture nonlinear relationships that traditional linear models often overlook. This chapter explores the application machine learning (ML) in research, focusing on their ability manage interactions multidimensional datasets. Nonlinear are central understanding behaviors, socioeconomic factors, psychological processes. Machine models, including decision trees, neural networks, random forests, support vector machines, provide a flexible framework for capturing these intricate patterns. begins by examining limitations introduces essential suited modeling. A discussion follows how automatically detect threshold effects, offering superior predictive power robustness against noise compared methods. also covers practical challenges model evaluation, validation, handling imbalanced data, emphasizing cross-validation performance metrics tailored nuances Practical recommendations offered researchers, highlighting balance between accuracy interpretability, ethical considerations, best practices communicating results diverse stakeholders. demonstrates while robust solutions modeling relationships, successful sciences requires careful attention quality, selection, considerations. holds transformative potential complex informing data-driven psychology, sociology, political policy-making.

Language: Английский

---

Optimizing the chemical removal of phosphorus for wastewater treatment: Insights from interpretable machine learning modeling with binary classification of elasticity and productivity

Resources Conservation and Recycling, Journal Year: 2025, Volume and Issue: 215, P. 108147 - 108147

Published: Jan. 29, 2025

Language: Английский

---

SBFL fault localization considering fault-proneness

Journal of Systems and Software, Journal Year: 2025, Volume and Issue: unknown, P. 112363 - 112363

Published: Feb. 1, 2025

---

Forecasting mental states in schizophrenia using digital phenotyping data

PLOS Digital Health, Journal Year: 2025, Volume and Issue: 4(2), P. e0000734 - e0000734

Published: Feb. 7, 2025

The promise of machine learning successfully exploiting digital phenotyping data to forecast mental states in psychiatric populations could greatly improve clinical practice. Previous research focused on binary classification and continuous regression, disregarding the often ordinal nature prediction targets derived from rating scales. In addition, health ratings typically show important class imbalance or skewness that need be accounted for when evaluating predictive performance. Besides it remains unclear which algorithm is best suited tasks, eXtreme Gradient Boosting (XGBoost) long short-term memory (LSTM) algorithms being 2 popular choices studies. CrossCheck dataset includes 6,364 state surveys using 4-point scales 23,551 days smartphone sensor contributed by patients with schizophrenia. We trained 120 models 10 (e.g., Calm, Depressed, Seeing things) passive tasks (ordinal classification) (XGBoost, LSTM) over 3 horizons (same day, next week). A majority regression performed significantly above baseline, macro-averaged mean absolute error values between 1.19 0.77, balanced accuracy 58% 73%, corresponds similar levels performance these metrics are scaled. Results also showed do not account (mean error, accuracy) systematically overestimated performance, XGBoost par better than LSTM models, a significant yet very small decrease was observed as horizon expanded. conclusion, properly imbalance, demonstrated comparable prevalent approach without losing valuable information self-reports, thus providing richer easier interpret predictions.

Language: Английский

---

Optimal Feature Extraction Technique for Sentiment Analysis of Product Reviews for Product Development

International Journal of Scientific Research in Computer Science Engineering and Information Technology, Journal Year: 2025, Volume and Issue: 11(1), P. 1702 - 1714

Published: Feb. 7, 2025

Consumer review sites, social media and micro-blogs carry a wealth of information on the general perspective, experience feedback that consumers have products. When there is high volume product reviews, it can be challenging to developers sift through make decision based consumers’ sentiments. Sentiment Analysis, branch Artificial Intelligence, assists in providing data help businesses understand customers’ desire track how brands goods are perceived. performing feature extraction, converts raw text input into machine learning compatible format. A strong set necessary order achieve prediction object classification accuracy. Identifying an optimal combination critical for increasing overall performance classification. In this research, we tackle problem by identifying extraction technique Analysis using feature-level analysis. N-gram, POS techniques lexicons Stanford CoreNLP, TextBlob, SentiWordNet different combinations examined. Multinomial Naïve Bayes, Lexicon Bayes + Unsupervised ensemble classifiers were modeled reviews positive, neutral negative classes thereby combination. We explored real datasets two products; car model known as “Nissan Sentra” mobile phone “Samsung Galaxy A12”. The MNB classifications was provided N-Gram, Part Speech TextBlob features while unsupervised VADER.

Language: Английский

---

Entropy-based fuzzy 1norm twin random vector functional link networks for binary class imbalance learning

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 145, P. 110201 - 110201

Published: Feb. 11, 2025

Language: Английский

---

Prediction models for treatment response in migraine: a systematic review and meta-analysis

The Journal of Headache and Pain, Journal Year: 2025, Volume and Issue: 26(1)

Published: Feb. 12, 2025

Migraine is a complex neurological disorder with significant clinical variability, posing challenges for effective management. Multiple treatments are available migraine, but individual responses vary widely, making accurate prediction crucial personalized care. This study aims to examine the use of statistical and machine learning models predict treatment response in migraine patients. A systematic review meta-analysis were conducted assess performance quality predictive response. Relevant studies identified from databases such as PubMed, Cochrane Register Controlled Trials, Embase, Web Science, up 30th November 2024. The risk bias was evaluated using PROBAST tool, adherence reporting standards assessed TRIPOD + AI checklist. After screening 1,927 documents, ten met inclusion criteria, six included quantitative synthesis. Key data extracted sample characteristics, intervention types, outcomes, modeling methods, metrics. pooled analysis area under curve (AUC) yielded value 0.86 (95% CI: 0.67–0.95), indicating good performance. However, generally had high bias, particularly domain, by tool. highlights potential predicting heterogeneity emphasize need caution interpretation. Future research should focus on developing high-quality, comprehensive, multicenter datasets, rigorous external validation, standardized guidelines like AI. Incorporating multimodal magnetic resonance imaging (MRI) data, exploring symptom-treatment interactions, establishing uniform methodologies outcome measures, size calculations, missing handling will enhance model reliability applicability, ultimately improving patient outcomes reducing healthcare burdens. PROSPERO, CRD42024621366.

Language: Английский

---