fastISM: performant<i>in silico</i>saturation mutagenesis for convolutional neural networks

Nair, Surag; Shrikumar, Avanti; Schreiber, Jacob; Kundaje, Anshul

doi:10.1093/bioinformatics/btac135

Cited by 13 publications

(6 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Attribution methods quantify the importances of individual nucleotides in the input sequences using forward- ( e.g. in silico mutagenesis 6, 19 ) or back-propagation 20, 21 . These importance scores can be the basis for further clustering of activating sub-sequences into PWMs 22 , which, as with filter visualization, can in turn be compared to known TF motifs for biological insights.…”

Section: Introductionmentioning

confidence: 99%

ExplaiNN: interpretable and transparent neural networks for genomics

Novakovsky

Fornés

Saraswat

et al. 2022

Preprint

View full text Add to dashboard Cite

Sequence-based deep learning models, particularly convolutional neural networks (CNNs), have shown superior performance on a wide range of genomic tasks. A key limitation of these models is the lack of interpretability, slowing their broad adoption by the genomics community. Current approaches to model interpretation do not readily reveal how a model makes predictions, can be computationally intensive, and depend on the implemented architecture. Here, we introduce ExplaiNN, an adaptation of neural additive models1 for genomic tasks wherein predictions are computed as a linear combination of multiple independent CNNs, each consisting of a single convolutional filter and fully connected layers. This approach brings together the expressivity of CNNs with the interpretability of linear models, providing global (cell state level) as well as local (individual sequence level) insights of the biological processes studied. We use ExplaiNN to predict transcription factor (TF) binding and chromatin accessibility states, demonstrating performance levels comparable to state-of-the-art methods, while providing a transparent view of the model’s predictions in a straightforward manner. Applied to de novo motif discovery, ExplaiNN detects equivalent motifs to those obtained from specialized algorithms across a range of datasets. Finally, we present ExplaiNN as a plug and play platform in which pre-trained TF binding models and annotated position weight matrices from reference databases can be combined in a simple framework. We expect that ExplaiNN will accelerate the adoption of deep learning by biological domain experts in their daily genomic sequence analyses.

show abstract

Section: Introductionmentioning

confidence: 99%

ExplaiNN: interpretable and transparent neural networks for genomics

Novakovsky

Fornés

Saraswat

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…TISM's strength especially comes through for long sequences (e.g., > 20kb), and therefore it is extremely useful to detect, extract, and compare regulatory motifs across sequences and tasks 4 . While not as accurate as FastISM 23 or Yuzu 24 (because these are not approximations), TISM, in contrast, is applicable to any network written in any code base, any number of sequences, and only requires a few lines of code to turn the model's gradient into TISM values.…”

Section: Discussionmentioning

confidence: 99%

“…Here, we show how one can very simply approximate ISM from the model’s gradient. Approximating ISM enables the analysis of both large sets of sequences and long sequences (e.g., >100kb) s. While not as accurate as FastISM 23 or Yuzu 24 (because these are not approximations), TISM is applicable to any type of network and only requires a few lines of code to turn the model’s gradient on the reference sequence into TISM, and also requires less computation time. We show that the majority of TISM (89%, >0.58) values correlates well above ISM values from different model initializations, suggesting that TISM is sufficient to understand the model’s learned regulatory grammar and predict effects of sequence variants across different loci.…”

Section: Discussionmentioning

confidence: 99%

Quick and effective approximation ofin silicosaturation mutagenesis experiments with first-order Taylor expansion

Sasse,

Chikina,

Mostafavi

2023

Preprint

View full text Add to dashboard Cite

To understand the decision process of genomic sequence-to-function models, various explainable AI algorithms have been proposed. These methods determine the importance of each nucleotide in a given input sequence to the model’s predictions, and enable discovery ofcisregulatory motif grammar for gene regulation. The most commonly applied method isin silicosaturation mutagenesis (ISM) because its per-nucleotide importance scores can be intuitively understood as the computational counterpart toin vivosaturation mutagenesis experiments. While ISM is highly interpretable, it is computationally challenging to perform, because it requires computing three forward passes for every nucleotide in the given input sequence; these computations add up when analyzing a large number of sequences, and become prohibitive as the length of the input sequences and size of the model grows. Here, we show how to use the first-order Taylor approximation for ISM, which reduces its computation cost to a single forward pass for an input sequence, placing its scalability on equal footing with gradient-based approximation methods such as “gradient-times-input”. We show that the Taylor ISM (TISM) approximation is robust across different model ablations, random initializations, training parameters, and data set sizes. We use our theoretical derivation to connect ISM with the gradient values and show how this approximation is related to a recently suggested correction of the model’s gradients.

show abstract

“…By contrast, GOPHER provides a one-stop-shop for data processing of peak-based classification and quantitative regression analysis, training with data augmentations, and comprehensive model evaluation. GOPHER incorporates many popular model interpretability tools, such as first-layer filter visualization, global importance analysis, and attribution methods, including in silico mutagenesis 54,55 , saliency maps 56 , integrated gradients 57 , and SmoothGrad 58 .…”

Section: Discussionmentioning

confidence: 99%

Evaluating deep learning for predicting epigenomic profiles

Toneyan

Tang

Koo

2022

Preprint

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

fastISM: performantin silicosaturation mutagenesis for convolutional neural networks

Cited by 13 publications

References 25 publications

ExplaiNN: interpretable and transparent neural networks for genomics

ExplaiNN: interpretable and transparent neural networks for genomics

Quick and effective approximation ofin silicosaturation mutagenesis experiments with first-order Taylor expansion

Evaluating deep learning for predicting epigenomic profiles

Contact Info

Product

Resources

About