jonstargaryen / jstructure Goto Github PK

• parse additional, interaction-specific information from the PLIP xml files

• chain filters is cumbersome or mere impossible for cases where you want to work with a container and functions to access the contained elements will return streams, but lose capabilities of the wrapping object
• maybe use custom stream implemenations

jstructure needs a sophisticated logging solution

• the stream-based approach on composing pairs etc seems really slow when compared to some classic implementation

• parse non-amino acid entries of PDB files such as nucleotides and hetatms
• store them in the data model
• create functions to retrieve specific groups

• creating proteins/chains/groups/atoms by hand is cumbersome

• constructors tends to are overloaded and hard to read/interpret

• e.g. the case when testing with synthetic data

• there could be a classic builder for structures

• currently the distinction between ATOM and HETATM records is flawed
• thus, the outputted file will never match the input

• currently the AminoAcid enum is clogged with static information (which is its job though)
• it would be nicer however, if there would be a set of commonly used .cif files provided as resource, so the protein parser does not necessarily depend on an active internet connection
• by doing so, atom names could be inferred of amino acids, nucleotides, water, ions and ligands could be resolved on-the-fly
• also the way amino acid names are currently resolved could be improved

• resolving AbstractFeatureProvider instances by the FeatureProviderRegistry works
• however, only when they scanned and requesting classes are located in the same package
• is it possible to fix this? or is the OSGi concept of custom class loaders needed?

• e.g. the case for 2w0l, chain A, resNum 95