FACS pipeline - Fast AMP Classification System

If you use this software in a publication please cite

FACS: antimicrobial peptide screening in genomes and metagenomes Celio Dias Santos-Junior, Shaojun Pan, Xing-Ming Zhao, Luis Pedro Coelho bioRxiv 2019.12.17.880385; doi: https://doi.org/10.1101/2019.12.17.880385

NOTE: This is still a work in progress and, while the results of the tool should be correct, we are still working on making FACS easier to install and use.

Fast AMP Classification System pipeline is a system created by Celio Dias Santos Jr. and Luis Pedro Coelho, from Fudan University (Shanghai). It is distributed under MIT license and represents a new way to prospect AMPs in natural environments using metagenomic data or genomic data to generate large datasets of antimicrobial peptides.

(A modified version of the Prodigal software is distributed with FACS under the GPL license)

Applications

FACS can be used in a wide-ranging of scenarios, such as screening for novel AMPs generating candidates to further testing and patenting, as well as, determination of microbiome quorum sensing mechanisms linking AMPs to health conditions or presence of diseases.

Install

The preferred installation method uses conda (in particular, using bioconda and conda-forge) to install all dependencies in a FACS-private environment, so that should be available on your system.

If it is, the installation should be simply a matter of getting the source code:

$ git clone https://github.com/FACS-Antimicrobial-Peptides-Prospection/FACS/

And executing the installation script:

$ bash install.sh

Examples

FACS supports gzipped inputs

To run FACS on peptides:

    bash FACS.sh --mode p \
        --fasta example_seqs/expep.faa.gz \
        --outfolder out_peptides \
        --outtag example -t 4

In this case, we use example_seqs/expep.faa.gz as input sequence. This should be an amino-acid FASTA file. The outputs will be written into a folder called out_peptides, and FACS will 4 threads. Adapt as needed and see the full list of arguments below.

To run FACS on contigs, use:

    bash FACS.sh --mode c \
        --fasta example_seqs/excontigs.fna.gz \
        --outfolder out_contigs \
        --outtag out_contigs \
        -t 4

In this example, we use the example file excontigs.fna.gz which is a FASTA file with nucleotide sequences, writing the output to out_contigs.

To run FACS on paired-end reads, use:

    bash FACS.sh --mode r \
        --fwd example_seqs/R1.fq.gz \
        --rev example_seqs/R2.fq.gz \
        --outfolder out_metag \
        --outtag example_metag \
        -t 4

The paired-end reads are given as paired files (here, example_seqs/R1.fq.gz and example_seqs/R2.fq.gz). If you only have single-end reads, you can omit the --rev argument.

To run FACS to get abundance profiles, you only need the forward reads file and a reference with peptide sequences. You can run abundance mode using two different types of references:

Fasta file with peptide sequences:

    bash FACS.sh --mode a \
        --fwd example_seqs/R1.fq.gz \
        --fasta example_seqs/ref.faa.gz \
        --outfolder out_abundance \
        --outtag example_abundance \
        -t 4

FACS output with predicted AMPs, it will be the $outtag.tsv.gz file generated. To use this file containing the peptide sequences and the probabilities associated to the prediction, you can run the following commmand:

    bash FACS.sh --mode a \
        --fwd example_seqs/R1.fq.gz \
        --ref out_metag/example_metag.tsv.gz
        --outfolder out_abundance_pred
        --outtag example_abundance_pred
        -t 4

Usage

There are few options to make the running of the program a bit customized and speed up process according to the settings of the system available.

Usage: FACS.sh --mode c/r/p/a [options]

Here's a guide for avaiable options. Defaults for each option are showed inside the
brackets: [default].

Basic options:

-h, --help            Show this help message

-m                     Mode of operation, type:
                       "c" to work with contigs,
                       "p" to predict AMPs directly from a peptides FASTA file,
                       "r" to work with reads, 
                       "a" to map reads against AMP output database and generate abundances table
	
--fasta               Compressed (or not gzipped) contigs or peptides fasta file

--fwd                 Illumina sequencing file in Fastq format (R1), please leave it
                      compressed and full path

--rev                 Illumina sequencing file in Fastq format (R2), please leave it
                      compressed and pass the full path

--ref                 Output of module "c" in its raw format [file type tsv and compressed]

--outfolder           Folder where output will be generated [Default: .]

--outtag              Tag used to name outputs [Default: FACS_OUT]

-t, --threads [N]     Number of threads [Default: 90% of available threads]

--block               Bucket size (take in mind it is measured in Bytes and also it 
                      influences memory usage). [Defeault: 100MB]

--log                 Log file name. FACS will save the run results to this log file
                      in output folder.

--mem                 Memory available to FACS ranging from 0 - 1. [Defult: 0.75]

--tmp                 Temporary folder (address)

--cls                 Cluster peptides: yes (1) or no (0). [Default: 1 - yes]

Pipeline overview

In general, FACS uses the pigz software to compress and decompress files quickly, as well as GNUParallel to make the shell processes faster and parallelized. The 22 descriptors adopted by FACS are hybrid comprising local and global contexts to do the sequence encoding. FACS performs firstly a distribution analysis of three classes of residues in two different features (Solvent accessibility and Free energy to transfer from water to lipophilic phase) as shown in Table 1. The novelty in this method is using the Free energy to transfer from water to lipophilic phase (FT) firstly described by Von Heijne and Blomberg, 1979 to capture the spontaneity of the conformational change that AMPs suffer while their transference from water to the membrane. For more info about the other descriptors used in FACS and the algorithms used to train the classifiers, please refer to the FACS reference.

Table 1. Classes adopted to the sequence encoding of the distribution at the first residue of each class. The Solvent Accessibility was adopted as in previous studies (Dubchak et al. 1995, 1999), however, the new feature FT was adapted from Von Heijne and Blomberg, 1979.

Properties	Class I	Class II	Class III
Solvent accessibility	A, L, F, C, G, I, V, W	R, K, Q, E, N, D	M, S, P, T, H, Y
FT	I,L,V,W,A,M,G,T	F,Y,S,Q,C,N	P,H,K,E,D,R

The other descriptors used in FACS classifiers are widely used in the AMPs description, as follows:

tinyAA (A + C + G + S + T)
smallAA (A + B + C + D + G + N + P + S + T + V)
aliphaticAA (A + I + L + V)
aromaticAA (F + H + W + Y)
nonpolarAA (A + C + F + G + I + L + M + P + V + W + Y)
polarAA (D + E + H + K + N + Q + R + S + T + Z)
chargedAA (B + D + E + H + K + R + Z)
basicAA (H + K + R)
acidicAA (B + D + E + Z)
charge (pH = 7, pKscale = "EMBOSS")
pI (pKscale = "EMBOSS")
aindex (relative volume occupied by aliphatic side chains - A, V, I, and L)
instaindex -> stability of a protein based on its amino acid composition
boman -> overall estimate of the potential of a peptide to bind to membranes or other proteins as receptor
hydrophobicity (scale = "KyteDoolittle") -> GRAVY index
hmoment (angle = 100, window = 11) -> quantitative measure of the amphiphilicity perpendicular to theaxis of any periodic peptide structure, such as the alpha-helix or beta-sheet

We opted to use random forests after some tests with alternative algorithms. The training of AMPs classifier used the training and validation data sets used by Bhadra et al. (2018), while the classifier of hemolytic peptides was trained and tested with the data sets previously established by Chaudhary et al. (2016).

The models here mentioned were implemented in an R script to filter off the non-AMP peptides and classify AMPs into hemolytic or not. After that, this script also submits the predicted AMPs to a decisions tree, classifying AMPs into 4 families:

Anionic linear peptides (ALP)
Anionic disulfide-bond forming peptides (ADP)
Cationic linear peptides (CLP)
Cationic disulfide-bond forming peptides (CDP)

Benchmark procedures showed that R22 models are efficient in retrieving AMPs with statistics that overcome the state-of-art methods (Table 2).

Table 2. Comparison of FACS and other state-of-art AMP prediction systems. All systems were tested with the benchmark data set from the AMPEP study available in Bhadra et al. (2018).

Model/Method	Acc.	Sp.	Sn.	Precision	MCC	Refererence
AMPep	0.962	0.965	0.95	0.913	0.9	doi:10.1038/s41598-018-19752-w
Peptide Scanner v2	0.755	0.686	0.943	0.523	0.557	doi: 10.1093/bioinformatics/bty179
CAMPr3-NN	0.728	0.704	0.794	0.494	0.445	doi: 10.1093/nar/gkv1051
CAMPr3-RF	0.584	0.461	0.923	0.384	0.354	doi: 10.1093/nar/gkv1051
CAMPr3-SVM	0.6	0.506	0.858	0.388	0.328	doi: 10.1093/nar/gkv1051
CAMPr3-DCA	0.617	0.542	0.821	0.396	0.324	doi: 10.1093/nar/gkv1051
iAMP	0.548	0.413	0.918	0.363	0.313	doi: 10.1038/srep42362
AMPA	0.779	0.941	0.336	0.675	0.361	doi: 10.1093/bioinformatics/btr604
R22_Ctrained	0.952	0.972	0.932	0.971	0.904	This study
R22_LargeTrainingset	0.967	1	0.934	1	0.936	This study

Meanwhile, the hemolytic prediction model implemented in FACS has a comparable performance of the state-of-art methods previously tested by Chaudhary et al., 2016 as shown in Table 3.

Table 3. Comparison of the performance of different hemolytic activity prediction systems. All the systems were trained and benchmarked with the same data sets (HemoPI) used by Chaudhary et al., 2016.

Methods	Study	Sn (%)	Sp (%)	Acc (%)	MCC
SVM	Chaudhary et al., 2016	95.7	94.8	95.3	0.91
IBK	Chaudhary et al., 2016	95.5	93.7	94.6	0.89
Multilayer Perceptron	Chaudhary et al., 2016	93.9	92.8	93.3	0.87
Logistic	Chaudhary et al., 2016	93.4	93.7	93.6	0.87
J48	Chaudhary et al., 2016	89.6	88.5	89.0	0.78
Random Forest	Chaudhary et al., 2016	94.1	94.6	94.3	0.89
Random Forest	This study	95.4	93.8	94.5	0.91

HiramHe / FACS