Generating and designing DNA with deep generative models

By Nathan Killoran, Leo J. Lee, Andrew Delong, David Kristjanson Duvenaud, Brendan J. Frey. 2017

165 citations recorded by the source corpus.

We propose generative neural network methods to generate DNA sequences and tune them to have desired properties. We present three approaches: creating synthetic DNA sequences using a generative adversarial network; a DNA-based variant of the activation maximization ("deep dream") design method; and a joint procedure which combines these two approaches together. We show that these tools capture important structures of the data and, when applied to designing probes for protein binding microarrays, allow us to generate new sequences whose properties are estimated to be superior to those found in the training data. We believe that these results open the door for applying deep generative models to advance genomics research.

Subfields: Artificial Intelligence, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Accelerating Protein Design Using Autoregressive Generative Models (2019, methodology): “Though others have reported difﬁculty training these models (Killoran et al., 2017), new machine learning techniques may make these models practical in the future (Kim et al., 2018; Yang et al., 2017; van den Oord & Vinyals, 2017).”
Conditioning by adaptive sampling for robust design (2019, methodology): “2 We compare our method, CbAS against the following methods: (1) AM-VAE —the activation-maximization method of Killoran et al. (2017).”
Likelihood Ratios for Out-of-Distribution Detection (2019, background): “…one of the popular choices for generating images (Oord et al., 2016; Van den Oord et al., 2016; Salimans et al., 2017) and sequence data such as genomics (Zou et al., 2018; Killoran et al., 2017) and drug molecules (Olive-crona et al., 2017; Gupta et al., 2018), and text (Jozefowicz et al., 2016).”
Machine learning applications in drug development (2019, background): “For instance, [51] have applied GANs to generate DNA sequences matching specific DNA motifs, that is, short DNA sequence variants which are associated with a specific function.”
A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications (2020, methodology): “Medical field: GANs have been widely utilized in medical field such as generating and designing DNA [404], [405], drug discovery [406], generating multi-label discrete patient records [407], medical image processing [408]–[415], dental restorations [416], and doctor recommendation [417].”
Feedback GAN (FBGAN) for DNA: a Novel Feedback-Loop Architecture for Optimizing Protein Functions (2018, background): “In independent and concurrent work, Killoran et al. use GANs to generate generic DNA sequences [13].”
PepCVAE: Semi-Supervised Targeted Design of Antimicrobial Peptide Sequences (2018, methodology): “The hydrophobic moment, a qualitative measure of the helical character within the sequence, is another frequently used descriptor used in cationic amphiphathic AMP classification Meher et al. [2017], Bhadra et al.”
Generative modeling for protein structures (2018, background): “Another result uses GANs to generate DNA sequences [24].”
DP-CGAN: Differentially Private Synthetic Data and Label Generation (2019, background): “DNNs have also been leveraged in health-related studies ranging from medical images [22, 39, 38, 21, 4] to human genome analyses [3, 29, 45].”
Deep Learning in Protein Structural Modeling and Design (2020, methodology): “data in predicted TF binding Killoran et al.(181)”
AdaLead: A simple and robust adaptive greedy search algorithm for sequence design (2020, methodology): “Killoran et al. (2017) pursue the sequence optimization by regularizing an “inverted model", φ (cid:48) θ − 1 ( y i ) = x i with a Wasserstein GAN (Arjovsky et al., 2017) which is trained to produce realistic samples.”