Merge pull request #211 from alexismhill3/paper-edits

Add JOSS editor revisions to Opfi paper

paper/paper.bib

@@ -99,7 +99,7 @@ @article{Blin:2019
 }
 
 @article{vanHeel:2018,
-    author = {van Heel, Auke J and de Jong, Anne and Song, Chunxu and Viel, Jakob H and Kok, Jan and Kuipers, Oscar P},
+    author = {{van Heel}, Auke J and {de Jong}, Anne and Song, Chunxu and Viel, Jakob H and Kok, Jan and Kuipers, Oscar P},
     title = "{BAGEL4: a user-friendly web server to thoroughly mine RiPPs and bacteriocins}",
     journal = {Nucleic Acids Research},
     volume = {46},

paper/paper.md

@@ -30,19 +30,19 @@ bibliography: paper.bib
 
 # Summary
 
-Gene clusters perform a diverse set of functions, many of which are relevant to biotechnology. There is a need for software tools that can extract candidate gene clusters from the vast amount of available genomic data. Therefore, we developed Opfi: a modular pipeline for identification of arbitrary gene clusters in assembled genomic or metagenomic sequences. Opfi contains functions for annotation, de-deduplication, and visualization of putative gene clusters. It utilizes a customizable rule-based filtering approach for selection of candidate systems that adhere to user-defined criteria. Opfi is implemented in Python, and is available on the Python Package Index and on Bioconda [@Grüning:2018].  
+Gene clusters are sets of co-localized, often contiguous genes that together perform specific functions, many of which are relevant to biotechnology. There is a need for software tools that can extract candidate gene clusters from vast amounts of available genomic data. Therefore, we developed Opfi: a modular pipeline for identification of arbitrary gene clusters in assembled genomic or metagenomic sequences. Opfi contains functions for annotation, de-deduplication, and visualization of putative gene clusters. It utilizes a customizable rule-based filtering approach for selection of candidate systems that adhere to user-defined criteria. Opfi is implemented in Python, and is available on the Python Package Index and on Bioconda [@Grüning:2018].  
 
 # Statement of need
 
-Gene clusters are sets of co-localized, often contiguous genes that together perform specific functions. These systems have been applied successfully to meet a number of biotechnological needs, including biofuel production, organic compound synthesis, and gene editing [@Fischbach:2010].  Although there are many tools available for annotation of singular genes (or protein domains) in biological sequence data [@Camacho:2009; @Steinegger:2017; @Buchfink:2021], these programs do not identify whole gene clusters out of the box. In many cases, researchers must combine bioinformatics tools ad-hoc, resulting in one-off pipelines that can be difficult to reproduce. Several software packages have been developed for discovery of specific gene clusters [@Blin:2019; @Santos-Aberturas:2019; @vanHeel:2018], but these tools may not be sufficiently flexible to identify clusters of an arbitrary genomic composition. To address these gaps, we developed a modular pipeline that integrates multiple bioinformatics tools, providing a flexible, uniform computational framework for identification of arbitrary gene clusters. In a recent study, we used Opfi to uncover novel CRISPR-associated transposons (CASTs) in a large metagenomics dataset [@Rybarski:2021].
+Gene clusters have been successfully repurposed for a number of biotechnical applications, including biofuel production, organic compound synthesis, and gene editing [@Fischbach:2010]. Despite the broad utility of known gene clusters, identification of novel gene clusters remains a challenging task. While there are many tools available for annotation of singular genes (or protein domains) in biological sequence data [@Camacho:2009; @Steinegger:2017; @Buchfink:2021], these programs do not identify whole gene clusters out of the box. In many cases, researchers must combine bioinformatics tools ad hoc, resulting in one-off pipelines that can be difficult to reproduce. Several software packages have been developed for the discovery of specific types of gene clusters [@Blin:2019; @Santos-Aberturas:2019; @vanHeel:2018], but these tools may not be sufficiently flexible to identify clusters of an arbitrary genomic composition. To address these gaps, we developed a modular pipeline that integrates multiple bioinformatics tools, providing a flexible, uniform computational framework for identification of arbitrary gene clusters. In a recent study, we used Opfi to uncover novel CRISPR-associated transposons (CASTs) in a large metagenomics dataset [@Rybarski:2021].
 
 # Implementation
 
 Opfi is implemented in Python, and uses several bioinformatics tools for feature annotation [@Camacho:2009; @Steinegger:2017; @Buchfink:2021; @Edgar:2007; @Shi:2019]. Users can install Opfi and all of its dependencies through Bioconda [@Grüning:2018]. Opfi consists of two major components: Gene Finder, for discovery of gene clusters, and Operon Analyzer, for rule-based filtering, deduplication, and visualization of gene clusters identified by Gene Finder. All modules generate output in a comma-separated (CSV) format that is common to the entire package.
 
 ## Example Gene Finder usage
 
-The following example script searches for putative CRISPR-Cas loci in the genome of *Rippkaea orientalis PCC 8802*. Information about the biological significance of this example, as well as data inputs and descriptions, can be found in the `tutorials` directory in the project GitHub repository. The example illustrates the use of the `Pipeline` class for setting up a gene cluster search. First, `add_seed_step` specifies a step to annotate *cas1* genes, using protein BLAST (BLASTP) [@Camacho:2009] and a database of representative Cas1 protein sequences. 10,000 bp regions directly up- and downstream of each putative *cas1* gene are selected for further analysis, and all other regions are discarded. Next, `add_filter_step` adds a step to annotate candidate regions for additonal *cas* genes. Candidates that do not have at least one additional *cas* gene are discarded from the master list of putative systems. Finally, `add_crispr_step` adds a step to search remaining candidates for CRISPR arrays, i.e. regions of alternatating ~30 bp direct repeat and variable sequences, using the PILER-CR repeat finding software [@Edgar:2007]. 
+The following example script searches for putative CRISPR-Cas loci in the genome of *Rippkaea orientalis PCC 8802*. Information about the biological significance of this example, as well as data inputs and descriptions, can be found in the `tutorials` directory in the project GitHub repository. The example illustrates the use of the `Pipeline` class for setting up a gene cluster search. First, `add_seed_step` specifies a step to annotate *cas1* genes, using protein BLAST (BLASTP) [@Camacho:2009] and a database of representative Cas1 protein sequences. 10,000 bp regions directly up- and downstream of each putative *cas1* gene are selected for further analysis, and all other regions are discarded. Next, `add_filter_step` adds a step to annotate candidate regions for additonal *cas* genes. Candidates that do not have at least one additional *cas* gene are discarded from the master list of putative systems. Finally, `add_crispr_step` adds a step to search remaining candidates for CRISPR arrays, i.e. regions of alternating ~30 bp direct repeat and variable sequences, using the PILER-CR repeat finding software [@Edgar:2007]. 
 
 ```python
 from gene_finder.pipeline import Pipeline
@@ -63,7 +63,7 @@ Running this code creates the CSV file `r_orientalis_results.csv`, which contain
 
 ## Example Operon Analyzer usage
 
-In the previous example, passing systems must meet the relatively permissive criterion of having at least one *cas1* gene co-localized with one additional *cas* gene. This is sufficient to identify CRISPR-Cas loci, but may also capture regions that do not contain functional CRISPR-Cas systems, but rather consist of open reading frames (ORFs) with weak homology to *cas* genes. These improbable systems could be eliminated during the homology search by making the match acceptance threshold more restrictive (i.e., by decreasing the e-value), however, this could result in the loss of interesting, highly diverged systems. Therefore, we implemented a module that enables post- homology search filtering of candidate systems, using flexible rules that can be combined to create sophisticated elimination functions. This allows the user to first perform a broad homology search with permissive parameters, and then apply rules to cull unlikely candidates without losing interesting and/or novel systems. Additionally, rules may be useful for selecting candidates with a specific genomic composition for downstream analysis. It should be noted that the use of the term "operon" throughout this library is an artifact from early development of Opfi. At this time, Opfi does not predict whether a candidate system represents a true operon, that is, a set of genes under the control of a single promoter. Although a candidate gene cluster may certainly qualify as an operon, it is currently up to the user to make that distinction. 
+In the previous example, passing systems must meet the relatively permissive criterion of having at least one *cas1* gene co-localized with one additional *cas* gene. This is sufficient to identify CRISPR-Cas loci, but may also capture regions that do not contain functional CRISPR-Cas systems, but rather consist of open reading frames (ORFs) with weak homology to *cas* genes. These improbable systems could be eliminated during the homology search by making the match acceptance threshold more restrictive (i.e., by decreasing the e-value), however, this could result in the loss of interesting, highly diverged systems. Therefore, we implemented a module that enables post-homology search filtering of candidate systems, using flexible rules that can be combined to create sophisticated elimination functions. This allows the user to first perform a broad homology search with permissive parameters, and then apply rules to cull unlikely candidates without losing interesting and/or novel systems. Additionally, rules may be useful for selecting candidates with a specific genomic composition for downstream analysis. It should be noted that the use of the term "operon" throughout this library is an artifact from early development of Opfi. At this time, Opfi does not predict whether a candidate system represents a true operon, that is, a set of genes under the control of a single promoter. Although a candidate gene cluster may certainly qualify as an operon, it is currently up to the user to make that distinction. 
 
 Rule-based filtering is illustrated with the following example. The sample script takes the output generated by the previous example and reconstructs each system as an `Operon` object. Next, the `RuleSet` class is used to assess each candidate; here, passing systems must contain two cascade genes (*cas5* and *cas7*) no more than 1000 bp apart, and at least one *cas3* (effector) gene. For a complete list of rules, see the Opfi [documentation](https://opfi.readthedocs.io/).