Macromolecular models

MacSyFinder relies on the definition of models of macromolecular systems as a set of models’ components to be searched by similarity search, and a set of rules regarding their genomic organization and their requirement level to make a complete system (mandatory, accessory components, number of components required).

See below for more details on MacSyFinder’s modelling scheme and the section on Functioning for the principles of the MacSyFinder’s search engine.

A MacSyFinder model (macsy-model for short) is the association of several elements:

  • a definition which describes the system to detect with a specific XML grammar that is described below.

  • a set of HMM profiles (one per component/gene in the model) to enable the similarity search of the systems’ components with the HMMER program.

The models are grouped by family possibly gathering sub-families (multiple levels allowed), for instance Secretion, Cas-proteins… A set of models from a same family (coherent set) of systems to detect is called hereafter a macsy-model package NEW in V2.

Principles, and how to write macsy-models definitions

Macsy-models are written as XML files, and should be named with the name of the system to detect as a prefix, and the XML file extension as a suffix. For example, ‘T1SS.xml’ for T1SS (Type I Secretion System).

A macsy-model defines a macromolecular System as:

  • A set of components (i.e. proteins, or protein-coding genes given the context) with different attributes that are used for system’s content description.

  • Features regarding the genomic architecture of the systems’ components for system detection.

  • Rules for quorum specifying how many components are required to infer the presence of a complete system.

Macsy-model Components

Four distinct types of components can be used to model the System’s content. Components correspond to Gene objects in MacSyFinder’s implementation, and point to corresponding HMM protein profiles.

  • mandatory components represent components that are essential to be found to infer the system’s presence.

  • accessory components correspond to components that can be found in some systems’ occurrence (or quickly evolving components that are hard to detect with a single HMM profile and thus can be missed along similarity search).

  • neutral components are used to build/extend clusters of proximal genes/components on the replicon analysed, but are not part of the quorum (i.e., not taken into account to assess the system’s presence). NEW in V2

  • forbidden components are components which presence is eliminatory for the system’s presence assessment.

Specifying a genomic organization

Beyond its list of Components, a MacSyFinder’s model of a System is defined by the genomic organization of its components. This genomic organization can be defined in several ways:

  • the general System’s architecture, whether it is single-locus or multi-loci (encoded at one or several loci)

  • the co-localization criteria defined either at the System level or at the Gene (component) level:

    • the inter-gene-max-space parameter (system- or gene- wise)

    • the loner parameter (gene- wise)

See below for more details on how to specify these parameters in a macsy-model.

The XML hierarchy

A System’s model is defined using a specific XML grammar that is hereby described. It consists in a hierarchic view of a Model that has specific features described through parameters, and is made of a set of Genes that have specific features themselves. All these elements and corresponding parameters will parametrize the search of Systems matching the search by MacSyFinder, in terms of Gene content and genomic architecture criteria.

  • The element root of a System’s model is “model”.

    • It has a mandatory attribute: “inter_gene_max_space”, an integer representing the maximal number of components without a match between two components with a match for a component profile in order to consider them contiguous (part of a same Cluster).

    • The version of the XML grammar (the actual version is “2.0”)

    • The element “model” may have attributes:

      • min_mandatory_genes_required: an integer representing the minimal number of mandatory genes required to infer the system’s presence.

      • min_genes_required: an integer representing the minimal number of mandatory or accessory genes (whose corresponding proteins match a profile of the model) required to infer the system’s presence.

      • multi_loci: a boolean set to True (“1”, “true” or “True”) to allow the definition of “scattered” systems (i.e., systems encoded at different genomic loci or by different gene clusters). If not specified, default value is false.

      • max_nb_genes define how many genes is necessary to consider a system as full. By default it is the sum of mandatory and accessory genes. But sometimes in special cases, there is 2 profiles, so 2 msf genes in model for one real gene. So in system only one gene can be detected and the wholeness is false.

    • The model contains one or more element(s) “gene” that correspond(s) to the genetic components of the macromolecular system.

  • The element “gene” has several mandatory attributes:

    • name: a string representing the name of the component/gene which must match that of a profile enclosed in the profile directory of the macsy-model package (see below).

    • presence: a string representing the status of the gene’s presence in the system. It can take four values among “mandatory”, “accessory”, “neutral”, “forbidden” (see above).

The element “gene” may have other attributes:

  • loner: a boolean. A loner gene can be isolated on the genome and does not have to be part of a cluster of genes to be considered for system’s assessment ( default false ).


    How loner works.

    A) The cluster 1 can be filled up with the loner D50 to reach the quorum defined in model A and form a system occurrence. B) There are 2 clusters and 2 loners (D50 and D60) and msf cannot assign which loner goes to which cluster. So msf picks the best loner (based on score) and sets the others as “counterpart”. 2 system occurrences are created whith the best loner. The user has to choose which loner hit can be assigned to which cluster. All loners found in the best solution are reported in best_solution_loners.tsv file. C) There are 2 clusters but only 1 loner. msf cannot decide to which cluster assign the loner. So the 2 system occurrences are proposed to the user in the output and a warning is raised to indicate the user should pick one. D) There are 2 clusters with one loner, but this loner is also multi_system. So the 2 clusters can be filled up with the loner.

  • multi_system: a boolean. If a gene has the feature “multi_system” (value set to “1”, “true” or “True”), it means that it can be used to fill multiple system occurrences (from a same model) - and thus be considered as part of several systems ( default false ).


    How multi_system works.

    A) The hit encoding for gene D in position 13 belongs to the system 1 (encoding model A). So it is used to fill up some other cluster, for instance cluster 2, which lacks this functionality. The cluster 2 then also fulfil the requirement of a system. B) The hit encoding for gene D in position 13 does not belong to a system. It cannot be used to fill up other clusters. In this example there is no system that satisfies the rules of model A. C) The gene D is present in the definition of model A and B. The hit encoding for gene D in position 13 belongs to the system 1 (encoding model A). It cannot be used to fill up the cluster 2 which codes for model B.

  • multi_model: a boolean. If a gene has the feature “multi_model” (value set to “1”, “true” or “True”), it means that two systems from different models can coexist in the best solution (they are said “compatible”) even if they share a component. The gene must be tagged as multi_model in both model definitions.


    How multi_model works.

    The hit encoding for gene D in position 13 is part of 2 systems: one for Model A, one for Model B. A) In both model definitions the gene D is tagged as multi_model. So the 2 systems can coexist in a same solution (they are “compatible”). B) The gene D is tagged as multi_model only in model A definition. The 2 systems are not compatible. So msf build 2 solutions and choose the best one. It has to be noted that this behaviour would actually be the same if gene D was not declared multi_model in either definitions.

  • inter_gene_max_space: an integer that defines gene-wise value of system’s “inter_gene_max_space” parameter (see above). It supersedes the system-wise parameter to give the gene a specific co-localization parameter.

The element “gene” may have one “exchangeables” child element:

  • The element “exchangeables” can contain one or more elements “gene”.

For a Gene to have “exchangeables” Genes listed, means that this Gene can be replaced in the quorum by the listed child Genes.


If the attributes inter_gene_max_space, loner, multi_model, multi_system are not specified for the exchangeable genes, then they inherit the values from the reference gene. Below some examples of attributes inheritance.

<gene name="A" presence="mandatory" multi_model="True">
        <gene name="B" />
        <gene name="C" />

In the snippet code above, the genes A/B/C are multi_model but not loner or multi_system.

<gene name="A" presence="mandatory">
        <gene name="B" multi_model="True"/>
        <gene name="C" />

In the snippet code above, The gene B is multi_model but not A and C.

<gene name="A" presence="mandatory" loner="True" multi_system="True">
        <gene name="B" />
        <gene name="C" multi_system="False"/>

In the snippet code above,

  • The genes A/B/C are loner

  • The genes A and B are multi_system, but not C.

<gene name="A" presence="mandatory" inter_gene_max_space="10">
        <gene name="B" inter_gene_max_space="5"/>
        <gene name="C" />

In the snippet code above, The genes A and C have an inter_gene_max_space = 10 whereas its value is 5 for the gene B .


The presence attribute is inevitably the same for the exchangeable genes than the reference gene.


If not specified by the user, several features will have their values assigned by default:

  • the genomic architecture of the System being searched will consist in a single locus. If a System may be made of Genes from multiple loci, consider setting the multi_loci parameter to True.

  • the quorum parameters min_mandatory_genes_required and min_genes_required will be set to the number of mandatory Genes listed - the accessory Genes being deemed not required to infer a complete System.

Example of a macsy-model definition in XML (more examples in our gallery of examples):

<model inter_gene_max_space="5" vers="2.0">
  <gene name="gspD" presence="mandatory">
         <gene name="sctC"/>
  <gene name="sctN_FLG" presence="mandatory" loner="1">
         <gene name="gspE"/>
         <gene name="pilT"/>
  <gene name="sctV_FLG" presence="mandatory"/>
  <gene name="flp" presence="accessory"/>

In this example, the described System consists of three mandatory and one accessory components:

  • Two components, the Gene “GspD” and the Gene “sctN_FLG” can respectively be replaced by sctC, and gspE and pilT genes in the quorum.

  • To be considered as part of such System, the components should be co-localized in loci (Clusters of Genes), which in this case would amount to being located from each other at a distance of 5-Genes maximum, except for the Gene “sctN_FLG” that is allowed to be located “alone” in the genome being investigated, by a loner parameter being set to True. As the multi_loci parameter is not set, by default the System should be made of a single locus (Cluster of co-localized Genes - except for the ones listed as loners).

  • To be considered a complete System, the quorum of Genes should be reached. In this case, the min_genes_required and min_mandatory_genes_required are not specified and therefore assigned to their default values: min_mandatory_genes_required is set to the number of mandatory Genes listed as well as the min_genes_required parameter (see above).


  • a gene is identified by its name.

  • this name is case sensitive.

  • this name must be unique inside a family of models.

  • a HMM profile with a gene-based name must exist in the profiles directory of the macsy-model package (see below).

Providing HMM profiles

For each gene mentioned in each model you have to provide a HMM profile to enable the similarity search of this gene. The HMM profile must have been created by the user from a curated multiple sequence alignment with the hmmbuild program from the HMMER package, or can have been obtained from HMM profiles’ databases such as TIGRFAM or PFAM .

This profile MUST have the same name as the name of the gene mentioned in the definition. For instance, a component named “GeneA” in the macsy-model would correspond by default to a HMM profile “GeneA.hmm” enclosed in the macsy-model package. The names are case-sensitive. All HMM profiles must be placed in the profiles directory of the macsy-model package.


For a detailed tutorial on how to define your macsy-model’s features, parameters and HMM profiles, you can have a look at our cookbook in this book chapter .