Final Report Summary - SISYPHE (Species Identity and SYmbiosis Formally and Experimentally explored)
SISYPHE (for “Species Identity and SYmbiosis f(PH)ormally and Experimentally explored”) had for main objective a mathematical and algorithmic study of symbiosis, which is commonly described as a close relationship between different biological species.
Symbiosis is a pervasive phenomenon, often of a long term nature. It was estimated that 50% of all known species are parasites, i.e. maintain a symbiotic relation with another species from which they benefit while the partner in the relation is harmed, and that close to a 100% of all plants and animals are parasitised as individuals. There is thus a growing recognition that symbiosis has a profound impact on the origin and maintenance of the biome and of its ecosystems, on health, and even on sex. Symbiosis thus appears essential to understand some of the most fundamental evolutionary and functional questions related to living organisms. Despite this, symbiotic relationships remain to this day little explored by computational biologists. By a highly pluri-disciplinary approach that blended mathematics and algorithmics with wet-lab experiments, the aim of SISYPHE was therefore to do an intensive, large-scale exploration of a variety of genomic and biochemical landscapes observed in the symbiont world, at the interface between symbionts and hosts, and of both with their environment.
This exploration focused initially on several aspects, namely genetic and metabolic dialogs, genome evolution, and symbiotic dynamics. The first two (genetic and metabolic dialogs) identify the main actors and describe by means of graph/network representations the interface between different organisms as concerns their gene expression regulation and metabolism. Genome evolution aims at identifying the changes to the genome composition and organisation of species involved in a symbiotic relation. Finally, symbiotic dynamics proposes to study the co-evolution of hosts and of their symbionts and to understand the genetic architecture of a parasitic invasion.
The computational study of biology in general, and of symbiosis more in particular, is fraught with the difficulties inherent to any science that is based on observations where data is most often missing, misleading, or possibly even erroneous, and where we clearly lack any precise picture of all the elements involved and of their role. Understanding in such cases must be progressively and tentatively reached through model testing.
In mathematical and algorithmic terms, this has a number of consequences: (i) some model(s) need to be proposed, (ii) those models need then to be exactly tested, (iii) the tests need to be exhaustive.
Because of a lack of knowledge, lack of data, and/or the complexity of what needs to be modelled, the first step may take a long time in maturing towards a reasonable yet useful model. It may also require refinement through successive model building and testing loops. All currently developed models remain thus clearly limited or inaccurate, whether this concerns metabolism – to study the interface between the networks of a host and its symbiont(s) or the impact of symbiosis on a host; gene expression – to evaluate the possible role of polymorphisms or small RNAs in the interaction with endosymbionts; genome evolution, to establish whether there is a difference between species in a harmful interaction (pathogens) and those in a positive or neutral relationship; or symbiotic dynamics – to evaluate the co-evolution of species and precisely establish the dynamics of their interactions as well as to understand the genetic architecture of a parasitic invasion.
The second step is often not considered essential, and indeed many of the most used methods for computational analysis of biological data are heuristics, i.e. do not guarantee any characteristic of the result obtained in relation to the solution sought. Without this however, it is clear that interpretation becomes difficult, and could be wrong. For instance, in the case of the transcriptome assembly of the insect Asobara tabida, the use of an available software built on several heuristics disregarded the polymorphism present in a gene called SXL. Instead, our exact method was able to recover such polymorphism and thus to suggest a new link between this gene and the observed phenotype.
Finally, as concerns the third step, even when exact solutions are sought, if more than one is possible, it is also often not considered essential to find them all. However, we showed within this project that this can be highly misleading. A most dramatic example of that is related to the study of the dynamics of species interactions. We showed, first that this highly depends on the estimated frequency of each of the possible co-evolutionary events that underlie such dynamics that is used as input for the method that then tries to establish the history of the interactions, and second, that there may be a huge number of scenarios for this history, all equally valid.
From an algorithmic perspective, this points to the importance of developing conceptual frameworks for exact enumeration techniques, or for techniques that can guarantee some characteristics of interest. This remains a largely underexplored area that has many other potential applications besides biology.
Symbiosis is a pervasive phenomenon, often of a long term nature. It was estimated that 50% of all known species are parasites, i.e. maintain a symbiotic relation with another species from which they benefit while the partner in the relation is harmed, and that close to a 100% of all plants and animals are parasitised as individuals. There is thus a growing recognition that symbiosis has a profound impact on the origin and maintenance of the biome and of its ecosystems, on health, and even on sex. Symbiosis thus appears essential to understand some of the most fundamental evolutionary and functional questions related to living organisms. Despite this, symbiotic relationships remain to this day little explored by computational biologists. By a highly pluri-disciplinary approach that blended mathematics and algorithmics with wet-lab experiments, the aim of SISYPHE was therefore to do an intensive, large-scale exploration of a variety of genomic and biochemical landscapes observed in the symbiont world, at the interface between symbionts and hosts, and of both with their environment.
This exploration focused initially on several aspects, namely genetic and metabolic dialogs, genome evolution, and symbiotic dynamics. The first two (genetic and metabolic dialogs) identify the main actors and describe by means of graph/network representations the interface between different organisms as concerns their gene expression regulation and metabolism. Genome evolution aims at identifying the changes to the genome composition and organisation of species involved in a symbiotic relation. Finally, symbiotic dynamics proposes to study the co-evolution of hosts and of their symbionts and to understand the genetic architecture of a parasitic invasion.
The computational study of biology in general, and of symbiosis more in particular, is fraught with the difficulties inherent to any science that is based on observations where data is most often missing, misleading, or possibly even erroneous, and where we clearly lack any precise picture of all the elements involved and of their role. Understanding in such cases must be progressively and tentatively reached through model testing.
In mathematical and algorithmic terms, this has a number of consequences: (i) some model(s) need to be proposed, (ii) those models need then to be exactly tested, (iii) the tests need to be exhaustive.
Because of a lack of knowledge, lack of data, and/or the complexity of what needs to be modelled, the first step may take a long time in maturing towards a reasonable yet useful model. It may also require refinement through successive model building and testing loops. All currently developed models remain thus clearly limited or inaccurate, whether this concerns metabolism – to study the interface between the networks of a host and its symbiont(s) or the impact of symbiosis on a host; gene expression – to evaluate the possible role of polymorphisms or small RNAs in the interaction with endosymbionts; genome evolution, to establish whether there is a difference between species in a harmful interaction (pathogens) and those in a positive or neutral relationship; or symbiotic dynamics – to evaluate the co-evolution of species and precisely establish the dynamics of their interactions as well as to understand the genetic architecture of a parasitic invasion.
The second step is often not considered essential, and indeed many of the most used methods for computational analysis of biological data are heuristics, i.e. do not guarantee any characteristic of the result obtained in relation to the solution sought. Without this however, it is clear that interpretation becomes difficult, and could be wrong. For instance, in the case of the transcriptome assembly of the insect Asobara tabida, the use of an available software built on several heuristics disregarded the polymorphism present in a gene called SXL. Instead, our exact method was able to recover such polymorphism and thus to suggest a new link between this gene and the observed phenotype.
Finally, as concerns the third step, even when exact solutions are sought, if more than one is possible, it is also often not considered essential to find them all. However, we showed within this project that this can be highly misleading. A most dramatic example of that is related to the study of the dynamics of species interactions. We showed, first that this highly depends on the estimated frequency of each of the possible co-evolutionary events that underlie such dynamics that is used as input for the method that then tries to establish the history of the interactions, and second, that there may be a huge number of scenarios for this history, all equally valid.
From an algorithmic perspective, this points to the importance of developing conceptual frameworks for exact enumeration techniques, or for techniques that can guarantee some characteristics of interest. This remains a largely underexplored area that has many other potential applications besides biology.