Designing Interpretable Convolution-Based Hybrid Networks for Genomics

Rs, Ghotra; Nk, Lee; Tripathy, Rohit; Koo, Peter K.

doi:10.1101/2021.07.13.452181

Cited by 5 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We optimized the model architecture, including kernel size and max pooling steps, to capture relevant genomic motifs (Fig. 5a) 90,91 . To identify potential cis-regulatory elements, we used the filter weights from the CNN module of APA-Net, which represent learned sequence motifs.…”

Section: Resultsmentioning

confidence: 99%

Single-nucleus transcriptome atlas of orbitofrontal cortex in amyotrophic lateral sclerosis with a deep learning-based decoding of alternative polyadenylation mechanisms

McKeever,

Sababi,

Sharma

et al. 2023

Preprint

View full text Add to dashboard Cite

The understanding of how different cell types contribute to amyotrophic lateral sclerosis (ALS) pathogenesis is limited. Here we generated a single-nucleus transcriptomic and epigenomic atlas of the frontal cortex of ALS cases with C9orf72 (C9) hexanucleotide repeat expansions and sporadic ALS (sALS). Our findings reveal shared pathways in C9-ALS and sALS, characterized by synaptic dysfunction in excitatory neurons and a disease-associated state in microglia. The disease subtypes diverge with loss of astrocyte homeostasis in C9-ALS, and a more substantial disturbance of inhibitory neurons in sALS. Leveraging high depth 3’-end sequencing, we found a widespread switch towards distal polyadenylation (PA) site usage across ALS subtypes relative to controls. To explore this differential alternative PA (APA), we developed APA-Net, a deep neural network model that uses transcript sequence and expression levels of RNA-binding proteins (RBPs) to predict cell-type specific APA usage and RBP interactions likely to regulate APA across disease subtypes.

show abstract

Section: Resultsmentioning

confidence: 99%

Single-nucleus transcriptome atlas of orbitofrontal cortex in amyotrophic lateral sclerosis with a deep learning-based decoding of alternative polyadenylation mechanisms

McKeever,

Sababi,

Sharma

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…First-layer convolutional filters provide an inductive bias to learn translational patterns such as motifs. However, the extent that they learn interpretable motifs largely depends on design choices, such as the max-pool size 40 , activation function 32 , or even the utilization of batch normalization 41 . However, it is not clear whether the same design principles established for binary models extends to quantitative models.…”

Section: Resultsmentioning

confidence: 99%

Evaluating deep learning for predicting epigenomic profiles

Toneyan

Tang

Koo

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Deep learning has been successful at predicting epigenomic profiles from DNA sequences. Most approaches frame this task as a binary classification relying on peak callers to define functional activity. Recently, quantitative models have emerged to directly predict the experimental coverage values as a regression. As new models continue to emerge with different architectures and training configurations, a major bottleneck is forming due to the lack of ability to fairly assess the novelty of proposed models and their utility for downstream biological discovery. Here we introduce a unified evaluation framework and use it to compare various binary and quantitative models trained to predict chromatin accessibility data. We highlight various modeling choices that affect generalization performance, including a downstream application of predicting variant effects. In addition, we introduce a robustness metric that can be used to enhance model selection and improve variant effect predictions. Our empirical study largely supports that quantitative modeling of epigenomic profiles leads to better generalizability and interpretability.

show abstract

“…4). Though-to our knowledge-this class of models is new to proteins, convolution-attention hybrid models have been used in genomics and found to serve as a sound inductive bias for discovering motifs and their interactions 27,28 .…”

Section: A Hybrid Neural-network Predicts Nucleationmentioning

confidence: 99%

Massive experimental quantification of amyloid nucleation allows interpretable deep learning of protein aggregation

Thompson,

Martín,

Olmo

et al. 2024

Preprint

View full text Add to dashboard Cite

Insoluble amyloid aggregates are the hallmarks of more than fifty human diseases, including the most common neurodegenerative disorders. The process by which soluble proteins nucleate to form amyloid fibrils is, however, quite poorly characterized. Relatively few sequences are known that form amyloids with high propensity and this data shortage likely limits our capacity to understand, predict, engineer, and prevent the formation of amyloid fibrils. Here we quantify the nucleation of amyloids at an unprecedented scale and use the data to train a deep learning model of amyloid nucleation. In total, we quantify the nucleation rates of >100,000 20-amino-acid-long peptides. This large and diverse dataset allows us to train CANYA, a convolution-attention hybrid neural network. CANYA is fast and outperforms existing methods with stable performance across diverse prediction tasks. Interpretability analyses reveal CANYA’s decision-making process and learned grammar, providing mechanistic insights into amyloid nucleation. Our results illustrate the power of massive experimental analysis of random sequence-spaces and provide an interpretable and robust neural network model to predict amyloid nucleation.

show abstract

Designing Interpretable Convolution-Based Hybrid Networks for Genomics

Cited by 5 publications

References 22 publications

Single-nucleus transcriptome atlas of orbitofrontal cortex in amyotrophic lateral sclerosis with a deep learning-based decoding of alternative polyadenylation mechanisms

Single-nucleus transcriptome atlas of orbitofrontal cortex in amyotrophic lateral sclerosis with a deep learning-based decoding of alternative polyadenylation mechanisms

Evaluating deep learning for predicting epigenomic profiles

Massive experimental quantification of amyloid nucleation allows interpretable deep learning of protein aggregation

Contact Info

Product

Resources

About