large language model

De novo generation of T-cell receptors with desired epitope-binding property by leveraging a pre-trained large language model

The paper introduces ERTransformer, a large language model that generates T-cell receptors with desired epitope-binding properties, showing high functional similarity to natural TCRs.

De novo generation of antibody CDRH3 with a pre-trained generative large language model

This paper introduces PALM, a pre-trained generative large language model for de novo antibody CDRH3 design, and A2binder for predicting antigen-antibody binding, showing effectiveness against SARS-CoV-2 variants.

GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT

GenePT embeds National Center for Biotechnology Information (NCBI) gene text descriptions using GPT-3.5 and aggregates them by expression to represent single cells without pretraining on millions of transcriptomes.

Bioinformatics Illustrations Decoded by ChatGPT: The Good, The Bad, and The Ugly

This study evaluates ChatGPT's ability to interpret bioinformatics illustrations, finding it effective for explaining plots and applying biological knowledge but lacking in quantitative analysis.

A General Single-Cell Analysis Framework via Conditional Diffusion Generative Models

This paper introduces a general single-cell analysis framework using conditional diffusion models, showing competitive performance across tasks with added flexibility via prior information.

THI1 Gene Evolutionary Trends: A Comprehensive Plant-Focused Assessment via Data Mining and Large-Scale Analysis

The paper analyzes the evolutionary trends of the THI1 gene across 702 homologs in Archaea and Eukarya, with a focus on plants, revealing conservation patterns and regulatory mechanisms.

Evolution of research topics and paradigms in plant sciences

The paper uses machine learning and language models to analyze plant science citations and track the evolution of research topics and paradigms over 50 years.

Cell2Sentence: Teaching Large Language Models the Language of Biology

Cell2Sentence converts single-cell gene expression profiles into text-like cell sentences so standard large language models like GPT-2 can be fine-tuned for cell generation and cell-type annotation.

Use of Large Language Models to Aid Analysis of Textual Data

This paper explores using large language models to assist in deductive coding of qualitative research texts, comparing LLM outputs to human coders and discussing implications and limitations.

xTrimoPGLM: Unified 100B-Scale Pre-trained Transformer for Deciphering the Language of Protein

A 100-billion-parameter protein language model trained with masked and autoregressive objectives together, topping prior results on 15 of 18 benchmarks and powering a fast structure predictor.

Efficient Visual Perception of Human-Robot Walking Environments Using Semi-Supervised Learning

This paper introduces a semi-supervised learning system for visual perception in human-robot walking environments, achieving high accuracy with fewer labeled images.

Automated Extraction and Visualization of Metabolic Networks from Biomedical Literature Using a Large Language Model

This study uses GPT-3.5-Turbo to extract and visualize metabolic networks from PubMed abstracts, generating a directed graph of metabolic interactions.

Machine learning prediction of enzyme optimum pH

This study develops and evaluates machine learning models to predict enzyme optimum pH, finding that language model embeddings significantly improve prediction accuracy, leading to the release of the EpHod model.

Literate programming for iterative design-build-test-learn cycles in bioengineering

The paper introduces teemi, an open-source platform for bioengineering that supports simulation, data integration, and machine learning for metabolic pathway design under FAIR principles.

Protein-Protein Interaction Prediction is Achievable with Large Language Models

This paper demonstrates that large language models can predict protein-protein interactions with high accuracy using synthetic data and fine-tuned models.

HistoMIL: a Python package for training Multiple Instance Learning models on histopathology slides

HistoMIL simplifies training Multiple Instance Learning models on histopathology slides, achieving up to 85% AUROC for cancer gene prediction.

Machine Learning-Guided Antibody Engineering That Leverages Domain Knowledge To Overcome The Small Data Problem

This paper presents a machine learning platform that enhances antibody affinity and neutralizing potency using domain knowledge and limited experimental data.

Efficient evolution of human antibodies from general protein language models and sequence information alone

Consensus likelihoods from general, unsupervised protein language models evolve seven human antibodies using only a wildtype sequence and <=20 tested variants, improving binding affinity up to 160-fold.

AI becomes a masterbrain scientist

This study explores GPT-4's ability to function as a masterbrain scientist in a closed-loop biological research process, generating hypotheses, designing experiments, analyzing results, and drawing conclusions.

The landscape of biomedical research

The paper creates a 2D map of biomedical literature to visualize research trends and evolution across disciplines.

Leveraging medical Twitter to build a visual-language foundation model for pathology AI

The paper builds a visual-language foundation model for pathology AI using a large dataset of pathology images and natural language descriptions curated from medical Twitter.

Deep Active Learning for Robust Biomedical Segmentation

This paper introduces a deep active learning framework that reduces annotation needs by 39-54% while improving robustness in medical image segmentation.

A publishing infrastructure for AI-assisted academic authoring

This paper explores using AI to improve scholarly writing by integrating large language models into a publishing ecosystem, showing they can enhance text clarity through revisions.

Language models generalize beyond natural proteins

A sequence-only protein language model, sampled under a structure-compatibility energy, produced designs of which 152 of 228 tested behaved as soluble monomers, some with no detectable UniRef90 match.

modelRxiv: A platform for the dissemination and interactive display of models

modelRxiv provides an interactive platform for sharing and exploring ecological and evolutionary models, enabling users to regenerate results with custom parameters.

X-MOL: large-scale pre-training for molecular understanding and diverse molecular analysis

X-MOL uses large-scale pre-training on 1.1 billion molecules to improve molecular analysis tasks through fine-tuning, achieving state-of-the-art results in property prediction, reaction analysis, and molecule generation.

Thinking ahead: prediction in context as a keystone of language in humans and machines

In nine epilepsy patients, cortical activity carries next-word information up to 800 ms before onset, tracks GPT-2 surprise after onset, and fits GPT-2 contextual embeddings better than static ones.

Designing Feature-Controlled Humanoid Antibody Discovery Libraries Using Generative Adversarial Networks

This paper uses a GAN to generate diverse human antibody libraries with controlled properties, validated through phage display and biophysical evaluation.

Ludicrous Speed Linear Mixed Models for Genome-Wide Association Studies

The paper introduces a cloud-optimized version of FaST-LMM for GWAS, enabling analysis of one million SNPs across one million individuals in two weeks with 868 CPU days.