The brainchaild of the Concept Web Alliance is the notion of ‘nanopublication’. Today, only a few years later, nanopublications are widely studied and accepted as a way to deal with massive data sets, interoperability of data, data citation and crowdsourcing. [...] The name nanopublication was coined for a computer readable, single meaningful assertion, including its rich provenance. Just like any classical publication, a nanopublication has authors, a time date stamp, a unique reference etc. The guidelines for nanopublications as now developed by the CWA recommend keeping the assertion in a nanopublication ‘as small as possible to be a self-contained scientific claim’. The smallest thinkable associative assertion between two concepts simply states in a Subject>Predicate>Object triple how two concepts relate to each other, for instance Protein A > interacts with > Protein B. In practice, many nanopublications need more than one triple in the assertional part. Also provenance and context (such as for instance the conditions under which a reaction takes place) can be formatted as triples in RDF or related computer readable languages.

Alltogether these triples form a small graph that represents the Nanopublication in computer readable format which can be represented to human users in their own language of choice at the same time. In order to create high quality nanopublications, each concept in the graph should be mapped to a unique reference (identifier) that unambiguously defines which concept is referred to. This major effort, which requires rich ontologies, Identity Mapping Services and many other software workflows and standards is an international effort by default and takes place in the context of the CWA and several international projects at the moment.

The period in which ‘web publishing’ apparently meant ‘putting dead PDF’s on the Internet’ is as good as over. However, the so called ‘article of the furture’, in fact not more than a mishappened Christmas tree of hyperlinks in the text that was originally meant for reading, is not the way to go. The mistake is that publishers are apparently as of yet still unable to think enough out-of-the-box to depart from the ‘article’ as the principle unit of scientific cmmunication and migrate towards entirely new and computer enabled ways of scholarly communication. The brief period in which journals started to accept ‘supplementary data’ linked to classical articles is also coming to an end. The first journals are already abandoning this policy. First, because reviewers hardly ever come around to review these data sets. Second, because the data are frequently at the proverbial ‘laptop of the PhD student’ and therefore untraceable after a few years and third, because most data are in exotic, undoubtedly brilliant, but unreadable formats reinvented on a daily basis by bioinformaticians. On top of all of this, the archaic system used by universities and funding agencies to judge scientific output based on the article as the smallest unit of scientific communication is rapidly becoming one of the single most inhibitory factors in eScience.

Ever since, approximately half of our research output is still on methodological improvements in biodata mining, disambiguation, and associative reasoning. However, the other half, and growing, is on actual knowledge discovery processes and new biological findings, only later to be confirmed in the laboratory.

However, in case you might think the demonstration of actual in silico knowledge discovery examples would silence the criticism, you are mistaken. In the review process of almost every article we submit, there is minimally one reviewer who writes a variant of the conservative and elitist statement: ‘I cannot accept that the computer is more clever than I am’. That is actually the last thing we suggest. We just demonstrate that the computer is much better than we are in systematically sifting through 22 million articles and thousands of databases.

The use of scientific literature and databases is rapidly shifting from ‘exploratory’ reading to ‘confirmational reading’ to check elements of the hypotheses that massive data analyses throw at us. This requires a very different role for scholarly communication and this is not the classical scientific article.

Each Knowlet has an ‘attraction’ to each other Knowlet with similar concepts in its ‘Object Cloud’, while some knowlets have very little to do with each other and may even repulse each other, some others may move closer and close in the Concept Web, pushed and pulled by many other knowlets in the space. The Knowlet of Malaria may move so close to the Knowlets of the cancer drug Tegafur that, even though they have never been mentioned together before, the Social Machine will hypothesise a potential anti-malaria role for the drug. [...] eScience needs different complementary levels of reasoning. Think of the metaphor of the helicopter view. One would never see the abbarent growth pattern in a cornfield caused by the remains of a Roman fortress when walking in the midst of the field. However, after spotting the pattern from the helicopter, one needs to land, take a shovel and dig to find the ruines. Next step would be the laboratory experiments to demonstrate the age of the stones before the conclusion can be drawn that indeed the pattern observed revealed a Roman fortress. Knowlets enable the helicopter view. With for instance Description Logics the immediate surroundings of the new associations can be explored (compared to the shovel), whilst final comfirmation of causal biological relationships in the wet Lab will follow.