#subtyping

Typesafe equality + generic lists?

Equality: equals(Object, Object) allows equals(2, "foo"). Typesafe equality says equals(T, T)

Generic Lists: equals(List, List)? Yes. equals(List, List)? No. But look out for covariance/contravariance.

Idea? List extends Eq<List<Eq>>

Problem! Undecidable. Inheritance Expansion and Nestred Contravariance.

Poison thing: Tree extends List

"This seems like a corner case"

List of equatables doesn't make sense to people... Eq<...> is not used as type arguments, parameters, etc...

So there seem to be two types of types. The shapes (Eq, Comparable) and materials (String, List). Empirically validated over 13.5 million lines.

Idea: Material-Shape Separation. Now, material inheritance is well-founded.

Decidable subtyping! Which is why, though subtyping is undecidable, this problem never happens in practice.

Key: expansion of shapes is limited. So, first reduce shapes, then reduce materials

Joins: String |_| Integer? ( Integer extends Clonable, etc... ) Limit join only to materials. Now computable.

mumble higher kinds

Also, conditional inheritance (which actually solves this problem)

Q: Are Shapes and Materials like Interfaces and Classes?

A: Orthogonal issue. Any type can be shape or mtaterial, it depends. Many of our examples are interfaces. But ther eare a lot of interfaces which do not use recursive inheritance. Trees as a list of tree, people never do that. As for predicate interface (indistinct)

Q: (Wadler) Felleisen, we did this years ago! Why didn't you cite our work? Isn't this what you did here? So it sounds like what you're arguing here is you should keep them separate.

A: I think we refer to typeclasses in the paper. For shapes, you just never pass around the program. You use them as constraints. "Shapes cause recursive inheritance"

For those who want to venture into Blastocystis subtyping - the easy way - I've recorded and uploaded a video on YouTube fyi.

To quote one of my colleagues attending the recent IWOP 2012 meeting in Tarrytown, NY, Blastocystis subtyping in humans and animals is becoming 'trendy', and so we keep trying to advocate for a standardisation of the metholodology of Blastocystis subtyping. We recently changed the title of our page at www.pubmlst.org/blastocystis so that now it is called Blastocystis Subtype (18S) and Sequence Typing (MLST) Databases, and we added some text to front page:

In terms of genetic markers, the barcode region (Scicluna et al., 2006) is by far the best represented in publicly available sequence databases, and the correct subtype can be identified by BLAST analysis in the sequence database at the present site. Blasting against this database has the added advantages, compared to using GenBank, of automatically assigning allele types to the SSU-rDNA as well as using the consensus subtype nomenclature (unlike GenBank where the subtype is included only if one was part of the accession submission and no attempt to impose a standard nomenclature is made). In case the sequence does not match any of the ones in the database despite full coverage of the region, this indicates that the sequence represents a new allele or maybe even a new subtype depending on the amount of variation. If a new subtype is suspected, we suggest doing PCR and sequencing of the complete SSU rRNA gene with subsequent phylogenetic analysis using reference sequences.

Now, the last bit is extremely important. We have seen examples of researchers (including ourselves!) assigning sequences to a new a subtype in the absence of complete SSU rDNA data (in fact complete sequences for ST10-ST14 are not yet publicly available!). Doing so has a least two major limitations/drawbacks: Far from all SSU rDNA regions have been validated as being representative of the whole SSU rRNA gene in terms of phylogenetic analysis, and therefore phylogenetic inferences based on non-validated regions may have little or at least less support than anticipated. Moreover, if someone analyses e.g. position 600-1600, and phylogenetic analysis based on this region reveals a potentially new subtype, this makes it impossible for his/her colleague who has data covering positions 1-600 from a Blastocystis isolate that may also represent a new subtype to ascertain whether it might be same subtype (see example below)! Obtaining complete SSU rDNA sequences directly from faecal DNA may be a cumbersome task but is sometimes possible by combining sequence-specific primers with low-specificity primers such as the RD5 and the RD3 primers (Clark, 1997). If a cultured isolate is available, obviously this makes complete SSU rDNA sequencing much easier. While it appears that the number of subtypes occurring in humans stays around 9, our gut feeling is that we are yet to uncover quite a few subtypes colonising non-human mammals, and it's great to see an increasing number of teams exploring the genetic diversity of Blastocystis. For instance, Dr Ronald Fayer and his group recently published exciting data on a new Blastocystis subtype in cattle, which they named ST14 (Fayer et al., 2012). Importantly, caution should be taken to avoid creating confusion in subtype terminology. Confusion can arise when independent researchers assign the same new subtype name (e.g. ST14, ST15, etc.) to novel sequences which in fact belong to different ribosomal lineages, or when incomplete SSU rDNA sequence data are used; this situation was seen recently, when Petrasova et al. (2011), assigned a Colobus sequence to ST5, although it was in fact a ST13 sequence (Clark et al., in press); the situation arose, since Petrasova et al. (2011) did not have data covering the region currently available for ST13 (Parkar et al., 2010), and therefore believed that their sequence was a unique ST5 variant. As for ST14, less than 500 bp are currently available, and these 500 bp are not in the barcode region, making it difficult for all teams using barcoding to compare their data. And so we would like to advocate for making complete SSU rDNA sequences publicly available (Genbank) for potentially new subtypes, for at least two reasons: 1. Phylogenetic inferences based on the complete SSU rDNA will be more robust than those obtained from analysing shorter sequence streches. 2. Complete seqeunces are needed for reference since subtype screening typically includes a single round PCR such as barcoding (Scicluna et al., 2006) amplifying about 550 bp; in the situation where complete SSU rDNAs are available for all known subtypes, it will be quick to analyse, whether a sequence may represent a new subtype, since this will be independent on the SSU rDNA region studied.We therefore hope that complete SSU rDNA sequences will soon be made publicly available for ST10-ST14. So, when does a complete SSU rDNA sequence represent a new subtype? Well, we have a review paper in press in Advances in Parasitology on recent developments in Blastocystis research, which will be published in less than six months probably, and which also touches on this topic; once the paper is published, I will try and make a summary our thoughts on this... Further reading:

Clark CG (1997). Extensive genetic diversity in Blastocystis hominis. Molecular and biochemical parasitology, 87 (1), 79-83 PMID: 9233675 Fayer R, Santin M, & Macarisin D (2012). Detection of concurrent infection of dairy cattle with Blastocystis, Cryptosporidium, Giardia, and Enterocytozoon by molecular and microscopic methods. Parasitology research PMID: 22710524 Parkar U, Traub RJ, Vitali S, Elliot A, Levecke B, Robertson I, Geurden T, Steele J, Drake B, & Thompson RC (2010). Molecular characterization of Blastocystis isolates from zoo animals and their animal-keepers. Veterinary parasitology, 169 (1-2), 8-17 PMID: 20089360

There has been some confusion about Blastocystis "subtypes" and "genotypes". 

Often, these two terms have been used interchangeably. While “subtype” refers to a distinct ribosomal lineage (which in the case of Blastocystis may very well be a distinct species), “genotype” denotes variation WITHIN subtypes. 

Currently, there is no clear definition of genotypes in Blastocystis. Based on phylogenetic analysis of barcode sequences of ST4, the existence of two genotypes in ST4 has been mentioned (Stensvold et al., 2011).  

Based on markers in the mitochondrion-like organelle of Blastocystis, we recently developed MLST assays for ST3 and ST4 and published data on intra-subtype variation in these two subtypes (Stensvold et al., 2012). While 58 sequence types (SQTs) were found among 81 ST3 isolates, only 5 SQTs were found among 50 ST4 isolates. 

By comparing SQTs with barcode sequences, we discovered that barcode sequences belonging to the same subtype may display intra-subtype diversity, and we found out that barcode sequences can be seen as valid proxies for SQTs. We have chosen to use the term "allele" to enable denotation of variation in barcode sequences. Currently, we have discovered 38 ST3 alleles (i.e. 38 different ST3 barcode sequences) as opposed to 8 different ST4 alleles. There are still no published data on ST1 and ST2 SQTs, but given the fact that 22 different alleles have been discovered so far for each of these two subtypes, we may expect a substantial number of SQTs.

The world of Blastocystis terminology and subtyping, etc. may seem a bit overwhelming and at times confusing, but believe me, - much has improved since 2006, when Blastocystis terminology was completely up in the air! 

For more information or further clarification, please don't hesitate to contact me.

Cited literature:

1. Stensvold CR, Alfellani M, Clark CG. Levels of genetic diversity vary dramatically between Blastocystis subtypes. Infect Genet Evol. 2012 Mar; 12 (2) :263-73. PubMed PMID:22116021.