Epistatic Net allows the sparse spectral regularization of deep neural networks for inferring fitness functions

Aghazadeh, Amirali; Nisonoff, Hunter; Ocal, Orhan; Brookes, David H.; Huang, Yalin; Koyluoglu, O. Ozan; Ramchandran, Kannan

doi:10.1038/s41467-021-25371-3

Cited by 28 publications

(25 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some algorithms specialize in modeling epistasis. Epistatic Net ( 27 ) introduced a neural network regularization strategy to limit the number of epistatic interactions. Other approaches focused on the global epistasis that arises due to a nonlinear transformation from a latent phenotype to the experimentally characterized function ( 28 , 29 ).…”

Section: Discussionmentioning

confidence: 99%

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

Gelman

Fahlberg

Heinzelman

et al. 2021

Proc. Natl. Acad. Sci. U.S.A.

146

View full text Add to dashboard Cite

The mapping from protein sequence to function is highly complex, making it challenging to predict how sequence changes will affect a protein’s behavior and properties. We present a supervised deep learning framework to learn the sequence–function mapping from deep mutational scanning data and make predictions for new, uncharacterized sequence variants. We test multiple neural network architectures, including a graph convolutional network that incorporates protein structure, to explore how a network’s internal representation affects its ability to learn the sequence–function mapping. Our supervised learning approach displays superior performance over physics-based and unsupervised prediction methods. We find that networks that capture nonlinear interactions and share parameters across sequence positions are important for learning the relationship between sequence and function. Further analysis of the trained models reveals the networks’ ability to learn biologically meaningful information about protein structure and mechanism. Finally, we demonstrate the models’ ability to navigate sequence space and design new proteins beyond the training set. We applied the protein G B1 domain (GB1) models to design a sequence that binds to immunoglobulin G with substantially higher affinity than wild-type GB1.

show abstract

Section: Discussionmentioning

confidence: 99%

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

Gelman

Fahlberg

Heinzelman

et al. 2021

Proc. Natl. Acad. Sci. U.S.A.

146

View full text Add to dashboard Cite

show abstract

“…Future work in this direction could focus on combining DAEs and RNNs for sequence based representation, and Graph Convolutional Networks (GCNs) for structure based as well as PPI based information. Combining these representations in a hierarchical classifier such as the multi-task DNN with biologically-relevant regularization methods 42 , 43 could allow for an explainable and computationally feasible DL architecture for protein function prediction.…”

Section: Major Successes Of DLmentioning

confidence: 99%

“…To make them human-interpretable, the input gets regularized using closed-form density functions of the data or GANs that mimic the data distribution. Methods that address the explainability question use more direct ways to gain insights from the NN function using their Taylor expansion 125 or Fourier transform 42 , 62 . The explanation takes the form of a heatmap which shows the importance of each input feature.…”

Section: General Challenges For DL In the Biosciencesmentioning

confidence: 99%

Current progress and open challenges for applying deep learning across the biosciences

et al. 2022

Self Cite

View full text Add to dashboard Cite

Deep Learning (DL) has recently enabled unprecedented advances in one of the grand challenges in computational biology: the half-century-old problem of protein structure prediction. In this paper we discuss recent advances, limitations, and future perspectives of DL on five broad areas: protein structure prediction, protein function prediction, genome engineering, systems biology and data integration, and phylogenetic inference. We discuss each application area and cover the main bottlenecks of DL approaches, such as training data, problem scope, and the ability to leverage existing DL architectures in new contexts. To conclude, we provide a summary of the subject-specific and general challenges for DL across the biosciences.

show abstract

“…For comparison we also fit an additive model using ordinary least squares and regularized pairwise and 3-way regression models. Since both L 1 and L 2 regularized regression have been used to infer genotype-phenotype maps [39,51,[64][65][66], here we fit the pairwise and three-way models using elastic net regression (see SI Appendix ) where the penalty term for model complexity is a mixture of L 1 and L 2 norms [67] and the relative weight of the two penalties is chosen via crossvalidation. In addition to the linear regression models, we also fit a global epistasis model [45] where the binding score is modeled as a nonlinear transformation of a latent additive phenotype on which each possible mutation has a background-independent e↵ect (SI Appendix ).…”

Section: Application To Protein Gmentioning

confidence: 99%

Higher-order epistasis and phenotypic prediction

Zhou

Wong

Chen

et al. 2020

Preprint

View full text Add to dashboard Cite

Contemporary high-throughput mutagenesis experiments are providing an increasingly detailed view of the complex patterns of genetic interaction that occur between multiple mutations within a single protein or regulatory element. By simultaneously measuring the effects of thousands of combinations of mutations, these experiments have revealed that the genotype-phenotype relationship typically reflects genetic interactions not only between pairs of sites, but also higher-order interactions between larger numbers of sites. However, modeling and understanding these higher-order interactions remains challenging. Here, we present a method for reconstructing sequence-to-function mappings from partially observed data that can accommodate all orders of genetic interaction. The main idea is to make predictions for unobserved genotypes that match the type and extent of epistasis found in the observed data. This information on the type and extent of epistasis can be extracted by considering how phenotypic correlations change as a function of mutational distance, which is equivalent to estimating the fraction of phenotypic variance due to each order of genetic interaction (additive, pairwise, three-way, etc.). We then infer the reconstructed sequence-function relationship using Gaussian process regression in which these variance components are used to define an empirical Bayes prior that in expectation matches the observed pattern of epistasis. To demonstrate the power of this approach, we present an application to the antibody-binding domain GB1 and provide a detailed exploration of a dataset consisting of high-throughput measurements for the splicing efficiency of human pre-mRNA 5′ splice sites for which we also validate our model predictions via additional low-throughput experiments.

show abstract

Epistatic Net allows the sparse spectral regularization of deep neural networks for inferring fitness functions

Cited by 28 publications

References 29 publications

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

Current progress and open challenges for applying deep learning across the biosciences

Higher-order epistasis and phenotypic prediction

Contact Info

Product

Resources

About