fastISM: performant in silico saturation mutagenesis for convolutional neural networks.
fastISM: performant in silico saturation mutagenesis for convolutional neural networks.
复制标题
fastISM：卷积神经网络的计算机模拟饱和诱变。
DOI：
10.1093/bioinformatics/btac135
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Kundaje,Anshul
中科院分区：
文献类型：
--
作者：
Nair,Surag;Shrikumar,Avanti;Schreiber,Jacob;Kundaje,Anshul
动机深度学习模型，如卷积神经网络，能够通过学习预测的新表征，准确地将生物序列映射到相关的功能读数和属性。在硅饱和诱变（ISM）是一种流行的特征归因技术，用于推断输入序列中所有字符对模型预测输出的贡献。ISM的主要缺点是它的运行时，因为它涉及通过训练模型对输入序列中每个字符的所有可能突变进行多次前向传播，以预测对输出的影响。结果提出了一种将常用卷积神经网络架构的ISM速度提高10倍以上的fasttism算法。fasttism是基于这样的观察：ISM中的大部分计算都花在卷积层上，单个突变只破坏中间层的有限区域，使大多数计算冗余。fasttism减少了基于反向传播的特征归属方法与ISM之间的差距。它远远超过了基于反向传播的方法在多输出体系结构上的运行时间，使得在大量序列上运行ISM成为可能。可用性和实现一个易于使用的Keras/TensorFlow 2实现的fasttism可以在https://github.com/kundajelab/fastISM上获得。fasttism可以使用pip install fasttism安装。可在https://colab.research.google.com/github/kundajelab/fastISM/blob/master/notebooks/colab/DeepSEA.ipynb.Supplementary information网站上找到动手教程。补充数据可在bioinformaticsonline上获得。
MotivationDeep-learning models, such as convolutional neural networks, are able to accurately map biological sequences to associated functional readouts and properties by learning predictivede novorepresentations.In silicosaturation mutagenesis (ISM) is a popular feature attribution technique for inferring contributions of all characters in an input sequence to the model’s predicted output. The main drawback of ISM is its runtime, as it involves multiple forward propagations of all possible mutations of each character in the input sequence through the trained model to predict the effects on the output.ResultsWe present fastISM, an algorithm that speeds up ISM by a factor of over 10× for commonly used convolutional neural network architectures. fastISM is based on the observations that the majority of computation in ISM is spent in convolutional layers, and a single mutation only disrupts a limited region of intermediate layers, rendering most computation redundant. fastISM reduces the gap between backpropagation-based feature attribution methods and ISM. It far surpasses the runtime of backpropagation-based methods on multi-output architectures, making it feasible to run ISM on a large number of sequences.Availability and implementationAn easy-to-use Keras/TensorFlow 2 implementation of fastISM is available at https://github.com/kundajelab/fastISM. fastISM can be installed using pip install fastism. A hands-on tutorial can be found at https://colab.research.google.com/github/kundajelab/fastISM/blob/master/notebooks/colab/DeepSEA.ipynb.Supplementary informationSupplementary data are available atBioinformaticsonline.