Morphology-Molecule Network (MMNet)

Introduction

Our main contribution is to propose a novel algorithm to improve the accuracy of species identification by combining molecular and morphological information.

Abstract

Integrative taxonomy is central to modern taxonomy and systematic biology, including behaviour, niche preference, distribution, morphological analysis and DNA barcoding. However, decades of use demonstrate that these methods can face challenges when used in isolation, for instance, potential misidentifications due to phenotypic plasticity for morphological methods, and incorrect identifications because of introgression, incomplete lineage sorting and horizontal gene transfer for DNA barcoding. Although researchers have advocated the use of integrative taxonomy, few detailed algorithms have been proposed. Here, we develop a convolutional neural network method (morphology-molecule network (MMNet)) that integrates morphological and molecular data for species identification. The newly proposed method (MMNet) worked better than four currently-available alternative methods when tested with 10 independent datasets representing varying genetic diversity from different taxa. High accuracies were achieved for all groups, including beetles (98.1% of 123 species), butterflies (98.8% of 24 species), fishes (96.3% of 214 species) and moths (96.4% of 150 total species). Further, MMNet demonstrated a high degree of accuracy (>98%) in four datasets including closely related species from the same genus. The average accuracy of two modest sub-genomic (single nucleotide polymorphism) datasets, comprising eight putative subspecies respectively, is 90%. Additional tests show that the success rate of species identification under this method most strongly depends on the amount of training data, and is robust to sequence length and image size. Analyses on the contribution of different data types (image versus gene) indicate that both morphological and genetic data are important to the model, and that genetic data contribute slightly more. The approaches developed here serve as a foundation for the future integration of multi-modal information for integrative taxonomy, such as image, audio, video, 3D scanning and biosensor data, to characterize organisms more comprehensively as a basis for improved investigation, monitoring and conservation of biodiversity.

Installation (You can install it through requirements.txt):

• Python 3.6.6 (Anaconda is recommended)
• torch (1.2.0)
• torchvision (0.3.0)
• visdom (0.1.8.8)

Data preprocess:

• Prepare an aligned FASTA file like:

  >BCIBT1645-13,Urbanus esmeraldus
  aacttt-atattttat-ttttggaat-ttgagcag...

  Note that the separator is a comma.

  >ID,species name

• Change the file name in DataPreprocess/preprocessFASTA2csv.py according to the FASTA file and run the script

• And the output file looks like:

idx	seqs	id	name	label
>FBCOB308-10,Adalia bipunctata	aacattatat…	FBCOB308-10	Adalia bipunctata	0
>COLFE1380-13,Adalia bipunctata	aacattatat…	COLFE1380-13	Adalia bipunctata	0
>COLFF981-13,Adalia bipunctata	aacattatat…	COLFF981-13	Adalia bipunctata	0
>COLFE416-12,Adalia bipunctata	aacattatat…	COLFE416-12	Adalia bipunctata	0
>FBCOB309-10,Adalia bipunctata	aacattatat…	FBCOB309-10	Adalia bipunctata	0
>FBCOF1053-12,Adalia decempunctata	aacattatat…	FBCOF1053-12	Adalia decempunctata	1
>COLFF267-13,Adalia decempunctata	aacattatat…	COLFF267-13	Adalia decempunctata	1
>COLFF266-13,Adalia decempunctata	aacattatat…	COLFF266-13	Adalia decempunctata	1
>COLFE085-12,Adalia decempunctata	aacattatat…	COLFE085-12	Adalia decempunctata	1
>FBCOA277-10,Adalia decempunctata	aacattatat…	FBCOA277-10	Adalia decempunctata	1
>COLFA615-12,Anatis ocellata	aacattatat…	COLFA615-12	Anatis ocellata	2
>COLFD685-12,Anatis ocellata	aacattatat…	COLFD685-12	Anatis ocellata	2
>COLFD686-12,Anatis ocellata	aacattatat…	COLFD686-12	Anatis ocellata	2
>COLFE438-12,Anatis ocellata	aacattatat…	COLFE438-12	Anatis ocellata	2
>COLFB032-12,Anatis ocellata	aacattatat…	COLFB032-12	Anatis ocellata	2

• Add the path of pictures in the rightmost column, then it looks like:

idx	seqs	id	name	label	img
>FBCOB308-10,Adalia bipunctata	aacattatat…	FBCOB308-10	Adalia bipunctata	0	/Path/To/Images/xxx.jpg
>COLFE1380-13,Adalia bipunctata	aacattatat…	COLFE1380-13	Adalia bipunctata	0	/Path/To/Images/xxx.jpg
>COLFF981-13,Adalia bipunctata	aacattatat…	COLFF981-13	Adalia bipunctata	0	/Path/To/Images/xxx.jpg
>COLFE416-12,Adalia bipunctata	aacattatat…	COLFE416-12	Adalia bipunctata	0	/Path/To/Images/xxx.jpg
>FBCOB309-10,Adalia bipunctata	aacattatat…	FBCOB309-10	Adalia bipunctata	0	/Path/To/Images/xxx.jpg
>FBCOF1053-12,Adalia decempunctata	aacattatat…	FBCOF1053-12	Adalia decempunctata	1	/Path/To/Images/xxx.jpg
>COLFF267-13,Adalia decempunctata	aacattatat…	COLFF267-13	Adalia decempunctata	1	/Path/To/Images/xxx.jpg
>COLFF266-13,Adalia decempunctata	aacattatat…	COLFF266-13	Adalia decempunctata	1	/Path/To/Images/xxx.jpg
>COLFE085-12,Adalia decempunctata	aacattatat…	COLFE085-12	Adalia decempunctata	1	/Path/To/Images/xxx.jpg
>FBCOA277-10,Adalia decempunctata	aacattatat…	FBCOA277-10	Adalia decempunctata	1	/Path/To/Images/xxx.jpg
>COLFA615-12,Anatis ocellata	aacattatat…	COLFA615-12	Anatis ocellata	2	/Path/To/Images/xxx.jpg
>COLFD685-12,Anatis ocellata	aacattatat…	COLFD685-12	Anatis ocellata	2	/Path/To/Images/xxx.jpg
>COLFD686-12,Anatis ocellata	aacattatat…	COLFD686-12	Anatis ocellata	2	/Path/To/Images/xxx.jpg
>COLFE438-12,Anatis ocellata	aacattatat…	COLFE438-12	Anatis ocellata	2	/Path/To/Images/xxx.jpg
>COLFB032-12,Anatis ocellata	aacattatat…	COLFB032-12	Anatis ocellata	2	/Path/To/Images/xxx.jpg

• Change the csv name in DataPreprocess/seq-img2database.py and run the script. The database of output could be found in the directory of Data.

Usage:

To test our algorithm, please follow these steps (tested on Ubuntu 16.04):

1 . Install the dependent packages

pip install -r requirements.txt

2 . Modify the main.py file in the directory (Set the path of training database and adjust hyperparameters according to your server). After you have modified the main.py file, please run the code as follows.

(Note: 0, 1, 2 use graphics cards No. 0, No. 1 and No. 2. If you run python train.sh directly, you use all graphics cards by default.)

CUDA_VISIBLE_DEVICES=0,1,2 python main.py

3 . The final result will be saved in the Output folder.

4 . Modify the directory of CalculateOptimizedThreshold.py then run this script. And you would get the dataset-specific threshold for species delimitation.

Citation

If you find our work helpful, please cite our paper:

@article{10.1093/sysbio/syab076,
    author = {Yang, Bing and Zhang, Zhenxin and Yang, Caiqing and Wang, Ying and Orr, Michael C and Hongbin, Wang and Zhang, Ai-Bing},
    title = "{Identification of Species by Combining Molecular and Morphological Data Using Convolutional Neural Networks}",
    journal = {Systematic Biology},
    year = {2021},
    month = {09},
    issn = {1063-5157},
    doi = {10.1093/sysbio/syab076},
    url = {https://doi.org/10.1093/sysbio/syab076},
    note = {syab076},
    eprint = {https://academic.oup.com/sysbio/advance-article-pdf/doi/10.1093/sysbio/syab076/40392247/syab076.pdf},
}