The study of global scientific co-authorship networks has been popular for several decades, and is still the subject of many academic papers (see here, here and here). Below are global co-authorship networks generated from the supplementary tables of the Publications Output: U.S. Trends and International Comparisons report which use relative cooperation indicators, that is to say: the relative tendency of a country to collaborate more with another country as compared to the global average.
What is the value of relative indicators? Because large countries like the USA and China have so much scientific output and research collaboration, they will almost always be placed at the heart of the global scientific co-authorship network in absolute terms. Hence, the relative index sheds light into the subtler tilts of countries in terms of their scientific collaboration.
Shown in the network graphs below are the relative co-authorship links of countries with collaborations with a frequency greater than 150% of the global average; the thickness of links represents the intensity of their collaboration in relative terms. Note that only countries with a certain research output threshold are included.
Co-authorship in 1996
Note the Asian and Pacific cluster, a Latin American cluster and a European cluster. The UK and US are connected to the Asian and Pacific cluster. There is also a (sub-)cluster with Egypt, Saudi Arabia, Pakistan and Turkey. Canada, Italy and Israel are not connected; their collaborations are globally balanced (none with a more than 150% global average frequency).
Co-authorship in 2020
In 2020, the European and South American clusters have morphed together, and Turkey has emerged as the main bridge between the Euro-Latin American cluster. Canada and the USA have a globally balanced network.
Bonus: Citations in 2020
The global citations network somewhat follows the co-authorship network, with a European and Latin American cluster and an Asian cluster. Singapore seems to form a bridge between both.