Biology subjects

Bileschi, M. L.

Publications and source records attributed to Bileschi, M. L..

5 recordsLinked to original sources

bioRxiv2026

A Vast Navigable Protein Sequence Subspace

Fitness landscapes govern evolution. Our empirical understanding of their structure has primarily been informed by evolutionary history. Here, we combine high-throughput DNA synthesis and machine learning, without training on evolutionary priors, to build a high-resolution map of the genome packaging landscape of the Adeno-associated Virus capsid protein. We assay 232,123 mutants around Adeno-Associated Virus 2 (AAV2) including 15 neighborhoods centered on seed sequences within 5-21 mutations of the wild type sequence. Local neighborhoods differ sharply in mutational tolerance and epistatic detail, yet much of the quantitative variation is captured by a simple pattern: sitewise effects transformed by global epistasis. Strikingly, using targeted sampling of only a few thousand mutants, we uncover a vast percolating network of function-preserving ridges embedded within a largely non-functional landscape. These ridges support billions of distinct trajectories of up to 667 steps that connect all sampled neighborhoods through functional single mutations. Designed variants with mutations distinct from the evolutionary background retain packaging at distances where randomly generated mutants rarely survive. These include capsids that combine up to 7 mutations that are largely absent from the evolutionary record, and which are each individually lethal in the 15 diverse backgrounds we tested. Our results reveal rich subspace structure resembling landscape orebodies: reservoirs of highly viable and navigable sequence space defined by simple constraints. These subspaces can be discovered through targeted sampling and can open access to vast troves of proteins that may be valuable for downstream applications such as therapeutics.

evolutionary biology↗

bioRxiv2024

ProtEx: A Retrieval-Augmented Approach for Protein Function Prediction

Mapping a protein sequence to its underlying biological function is a critical problem of increasing importance in biology. In this work, we propose ProtEx, a retrieval-augmented approach for protein function prediction that leverages exemplars from a database to improve accuracy and robustness and enable generalization to unseen classes. Our approach relies on a novel multi-sequence pretraining task, and a fine-tuning strategy that effectively conditions predictions on retrieved exemplars. Our method achieves state-of-the-art results across multiple datasets and settings for predicting Enzyme Commission (EC) numbers, Gene Ontology (GO) terms, and Pfam families. Our ablations and analysis highlight the impact of conditioning predictions on exemplar sequences, especially for classes and sequences less well represented in the training data.

bioinformatics↗

bioRxiv2021

ProteInfer: deep networks for protein functional inference

Predicting the function of a protein from its amino acid sequence is a long-standing challenge in bioinformatics. Traditional approaches use sequence alignment to compare a query sequence either to thousands of models of protein families or to large databases of individual protein sequences. Here we instead employ deep convolutional neural networks to directly predict a variety of protein functions - EC numbers and GO terms - directly from an unaligned amino acid sequence. This approach provides precise predictions which complement alignment-based methods, and the computational efficiency of a single neural network permits novel and lightweight software interfaces, which we demonstrate with an in-browser graphical interface for protein function prediction in which all computation is performed on the users personal computer with no data uploaded to remote servers. Moreover, these models place full-length amino acid sequences into a generalised functional space, facilitating downstream analysis and interpretation. To read the interactive version of this paper, please visit https://google-research.github.io/proteinfer/ O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/461077v2_ufig1.gif" ALT="Figure 1"> View larger version (32K): org.highwire.dtl.DTLVardef@74f8f1org.highwire.dtl.DTLVardef@183ae0borg.highwire.dtl.DTLVardef@1754f3org.highwire.dtl.DTLVardef@1ca54f5_HPS_FORMAT_FIGEXP M_FIG C_FIG QR code for the interactive version of this preprint at https://google-research.github.io/proteinfer/

bioinformatics↗

bioRxiv2019

Critiquing Protein Family Classification Models Using Sufficient Input Subsets

In many application domains, neural networks are highly accurate and have been deployed at large scale. However, users often do not have good tools for understanding how these models arrive at their predictions. This has hindered adoption in fields such as the life and medical sciences, where researchers require that models base their decisions on underlying biological phenomena rather than peculiarities of the dataset introduced. In response, we propose a set of methods for critiquing deep learning models and demonstrate their application for protein family classification, a task for which high-accuracy models have considerable potential impact. Our methods extend the sufficient input subsets technique, which we use to identify subsets of features (SIS) in each protein sequence that are alone sufficient for classification. Our suite of tools analyzes these subsets to shed light on the decision-making criteria employed by models trained on this task. These tools expose that while deep models may perform classification for biologically-relevant reasons, their behavior varies considerably across choice of network architecture and parameter initialization. While the techniques that we develop are specific to the protein sequence classification task, the approach taken generalizes to a broad set of scientific contexts in which model interpretability is essential.

bioinformatics↗

bioRxiv2019

Using Deep Learning to Annotate the Protein Universe

Understanding the relationship between amino acid sequence and protein function is a long-standing problem in molecular biology with far-reaching scientific implications. Despite six decades of progress, state-of-the-art techniques cannot annotate 1/3 of microbial protein sequences, hampering our ability to exploit sequences collected from diverse organisms. In this paper, we explore an alternative methodology based on deep learning that learns the relationship between unaligned amino acid sequences and their functional annotations across all 17929 families of the Pfam database. Using the Pfam seed sequences we establish rigorous benchmark assessments that use both random and clustered data splits to control for potentially confounding sequence similarities between train and test sequences. Using Pfam full, we report convolutional networks that are significantly more accurate and computationally efficient than BLASTp, while learning sequence features such as structural disorder and transmembrane helices. Our model co-locates sequences from unseen families in embedding space, allowing sequences from novel families to be accurately annotated. These results suggest deep learning models will be a core component of future protein function prediction tools.

bioinformatics↗