Identifying Evolutionary Origins of Repeat Domains in Protein Families

Improved inference of tandem domain duplications

“…Other models have considered horizontal transfers, merges/splits and shuffling events, but again with each duplication consisting of a single domain ( Hallett et al , 2004 ; Stolzer et al , 2015 ; Yi-Chieh et al , 2012 ). In contrast, Aluru and Singh (2020 ) and Dondi et al (2019 ) consider reconcilations where single duplication events can consist of multiple units (e.g. domains or sequences), but they do not consider the positional information of these units and thus these duplications are not constrained to be tandem duplications; these formulations, which we term as Concurrent Duplication Loss (CDL) reconciliations, can be defined using all conditions except Condition 3c from Definition 1, and with no eligibility matrix as input.…”

“…In Dondi et al (2019 ), a maximum parsimony approach with fixed duplication and loss costs is used, it is shown that the CDL reconciliation problem is NP-complete, and a fixed-parameter tractable algorithm (along with its implementation MultRec) is given. In Aluru and Singh (2020 ), more general duplication cost functions based on duplication size are considered, and an ILP solution is given. While these previous models did not explicitly consider tandem duplications, a key insight of our work is that we can define TDL reconciliations with a simple adaptation of the definition of CDL reconciliations ( Aluru and Singh, 2020 ; Dondi et al , 2019 ) with the use of an eligibility matrix, which specifies which domain pairs can be part of the same tandem duplication.…”

“…In Aluru and Singh (2020 ), more general duplication cost functions based on duplication size are considered, and an ILP solution is given. While these previous models did not explicitly consider tandem duplications, a key insight of our work is that we can define TDL reconciliations with a simple adaptation of the definition of CDL reconciliations ( Aluru and Singh, 2020 ; Dondi et al , 2019 ) with the use of an eligibility matrix, which specifies which domain pairs can be part of the same tandem duplication. In the next section, we show how the elements of this key matrix are computed from input domain and gene trees and the relative positions of domains in modern day sequences.…”

“…Once the eligibility matrix is computed, the TDL reconciliation problem can be solved with a simple modification to the ILP we previously introduced ( Aluru and Singh, 2020 ) to solve the CDL reconciliation problem. Solving the TDL reconciliation problem with this ILP requires the addition of a single constraint (see Supplementary Material for the full ILP solution).…”

“…Instead duplications are assumed to occur one domain at a time. In other approaches, multidomain duplications are explicitly included in the model ( Aluru and Singh, 2020 ; Bansal and Eulenstein, 2008 ; Dondi et al , 2019 ; Guigo et al , 1996 ). However, these approaches do not consider tandem duplications.…”