Efficient DNA sequence compression with neural networks

Abstract

Background: The increasing production of genomic data has led to an intensified need for models that can cope efficiently with the lossless compression of DNA sequences. Important applications include long-term storage and compression-based data analysis. In the literature, only a few recent articles propose the use of neural networks for DNA sequence compression. However, they fall short when compared with specific DNA compression tools, such as GeCo2. This limitation is due to the absence of models specifically designed for DNA sequences. In this work, we combine the power of neural networks with specific DNA models. For this purpose, we created GeCo3, a new genomic sequence compressor that uses neural networks for mixing multiple context and substitution-tolerant context models.

Findings: We benchmark GeCo3 as a reference-free DNA compressor in 5 datasets, including a balanced and comprehensive dataset of DNA sequences, the Y-chromosome and human mitogenome, 2 compilations of archaeal and virus genomes, 4 whole genomes, and 2 collections of FASTQ data of a human virome and ancient DNA. GeCo3 achieves a solid improvement in compression over the previous version (GeCo2) of $2.4\%$, $7.1\%$, $6.1\%$, $5.8\%$, and $6.0\%$, respectively. To test its performance as a reference-based DNA compressor, we benchmark GeCo3 in 4 datasets constituted by the pairwise compression of the chromosomes of the genomes of several primates. GeCo3 improves the compression in $12.4\%$, $11.7\%$, $10.8\%$, and $10.1\%$ over the state of the art. The cost of this compression improvement is some additional computational time (1.7-3 times slower than GeCo2). The RAM use is constant, and the tool scales efficiently, independently of the sequence size. Overall, these values outperform the state of the art.

Conclusions: GeCo3 is a genomic sequence compressor with a neural network mixing approach that provides additional gains over top specific genomic compressors. The proposed mixing method is portable, requiring only the probabilities of the models as inputs, providing easy adaptation to other data compressors or compression-based data analysis tools. GeCo3 is released under GPLv3 and is available for free download at https://github.com/cobilab/geco3.

Keywords: DNA sequence compression; context mixing; lossless data compression; mixture of experts; neural networks.

Figure 1:

Mixer architecture: (a) High-level overview of inputs to the neural network (mixer) used in GeCo3. Model₁ through Model_i represent the GeCo2 model outputs (probabilities for A, C, T, G). Perf represents the performance metrics (hit, best, bits) for each model. Freqs are the frequencies for the last 8, 16, and 64 symbols. nnbits is a moving average of the approximate number of bits that the neural network is producing. The network outputs represent the non-normalized probabilities for each DNA symbol. (b) A fully connected neural network with 1 hidden layer. For illustration purposes, this neural network only has the inputs corresponding to 1 model and the 3 features that evaluate the model performance. The frequencies of the last 8, 16, and 64 symbols, as well as the nnbits and the bias neurons, are omitted.

Figure 2:

Number of bytes (s) and time (t) according to the number of hidden nodes for the reference-free compression of ScPo, EnIn, and DrMe sequence genomes.

Figure 3:

Relative ratio and cost of GeCo3 compared with NAF and GeCo2 for sequences in DS1 and DS2. Higher relative ratios represent greater compression improvements by GeCo3. The cost is calculated assuming €0.13 per GB and the storage of 3 copies. The red dashed line shows the cost threshold. Cost points above the line indicate that GeCo3 is more expensive. Denisova32h represents the results of running the Denisova sequence with 32 instead of 64 hidden nodes.

Figure 4:

Comparison of histograms using the EnIn (Entamoeba invadens) and OrSa (Oryza sativa) genome sequences and GeCo2 and GeCo3 as data compressors.

Figure 5:

Complexity profile using the smoothed number of bits per symbol (Bps) of GeCo2 subtracted by GeCo3 Bps. The Bps were obtained by referential compression of PT_Y (Chromosome Y from Pan troglodytes) with the corresponding Homo sapiens chromosome, with the same parameters as in Table 4. Regions where the line rises above zero indicate that GeCo3 compresses more than GeCo2.