We’ve been waiting for ancient DNA to answer some questions about eastern Eurasia for a while. I always thought Qiaomei Fu would spearhead it, but it doesn’t seem like it worked out that way. That’s because she’s not on a new preprint, The Genomic Formation of Human Populations in East Asia, which fills in a lot of gaps and confusing aspects of what has been reported from fragments of publications that came before (e.g., this clarifies a lot of things with Japan, see below). Since there has already been ancient DNA work on eastern Siberia and Southeast Asia, this is really focusing on the area in and around what is today the Peoples’ Republic of China. The first author has an affiliation with a university in Fujian, a province in southeast China.
Much of the analysis can be understood as organized around language families, and the demographics associated with them. In this way, it goes back to L. L. Cavalli-Sforza’s correlations between gene trees and language trees, as well as his later work on the agricultural Diasporas.
First, there isn’t something radically surprising here in their results. As I suggest above, the mass of ancient DNA in the preprint and model-building just snap together a lot of what you can see in other work, some going back decades.
Let’s start with the “Onge-like/related ancestry. ”
Below you see the strange pattern of Y chromosomal haplogroup D. It’s common in Tibet, Japan, and among the Andamanese.
In the preprint, the authors argue that there is a deep division among East Eurasian populations, going back further than 40,000 years, between a set of populations descended from groups related to Tianyuan man, and populations with affinities to the indigenous peoples of southeast Eurasia and Australia (“Ancestral Ancestral South Indians”, AASI, the Onge, the Negritos of Malaysia and the Phillippines, and Oceanians). Modern populations in East Asia can be thought of as a mix between these two groups, in various pulses and waves. The finding that some peoples in the Amazon had “Australo-Melanesian” affinity is very strange, but note that there’s no guarantee that the geographic distribution of the two clades was so skewed in the past in a north-south manner.
The Onge-related ancestry is apparently found as the deepest layer in the Tibetan plateau and contributes 45% of the ancestry to the Jomon of Japan. Among ancient proto-Austronesian peoples of Taiwan, it contributed 14% of the ancestry. Earlier work on Southeast Asia indicated that even before the expansion of Austro-Asiatic farmers out of southern China they mixed with a basal East Eurasian lineage related to the Onge.
Chinese annals record the presence of dark-skinned peoples in Yunnan nearly into historical periods. These could very well be legends or rumors, or, they could be the last relic populations that had not been fully absorbed into the Tianyuan-descended farmer expansion.
Moving more recently into the past, the preprint findings that of the Tianyuan descended populations in East Asia there is a northern and southern grouping. The northern grouping has been discussed before, it is the classic Amur-river valley population. It turns out that a sample from 5,000 years ago in northern Shaanxi, just to the north of the hearth of classical 
Chinese civilization in Henan, resembles these Amur-river valley populations. Though the authors don’t have samples from southern China, or even the Yangzi, they use modern samples from southern Chinese peoples, as well as ancient samples from Taiwan, to infer that it is likely that the Yangzi river valley was inhabited by a somewhat different group during prehistory than the modern Han Chinese.
In the preprint, the argument is made that Austronesian, Tai-Kadai, and Austro-Asiatic all emerged out of the Yangzi valley and its rice cultures. As noted above, other papers have already outlined the peopling of Southeast Asia using ancient DNA, so I will ignore that. But, note that for Austro-Asiatic populations, ~1/3 of the ancestry is Onge-related. Some of this was mixed in while in southern China, but some of it probably accrued later on in Southeast Asia.
Modern Austro-Asiatic populations can then be thought of as a compound of Tianyuan, and various Onge-related groups.
China: