Wednesday, January 27, 2021

Yuri Milner interviews Donaldson, Kontsevich, Lurie, Tao, and Taylor (2015 Breakthrough Prize)


I came across this panel discussion recently, with Yuri Milner (former theoretical physicist, internet billionaire, and sponsor of the Breakthrough Prize) as interlocutor and panelists Simon Donaldson, Maxim Kontsevich, Jacob Lurie, Terence Tao, and Richard Taylor. 

Among the topics covered: the nature of mathematics, the simulation question, AGI and automated proof, human-machine collaboration in mathematics. Kontsevich marvels at the crystalline form of quantum mechanics: why linearity? why a vector space structure? 

Highly recommended!

See also 

The Quantum Simulation Hypothesis: Do we live in a quantum multiverse simulation? 

Sunday, January 24, 2021

Clinical Applications of Polygenic Risk Scores

Last week we posted a new paper (see bottom), prepared for the book Genomic Prediction of Complex Traits, Springer Nature series Methods in Molecular Biology. Someone asked me to comment more on clinical applications of polygenic risk scores. Here's what we say in the paper, using the specific example of breast cancer (emphasis added):
There is already signifcant interest in the application of PRS in a clinical setting, for example to identify high risk individuals who might receive early screening or preventative care [2, 13–24]. As a concrete example, women with high PRS scores for breast cancer can be offered early screening: already standard of care for those with BRCA risk variants [25, 26]. However, BRCA mutations affect no more than a few women per thousand in the general population [27–29]. Importantly, the number of (BRCA negative) women who are at high risk for breast cancer due to polygenic effects is an order of magnitude larger than the population of BRCA carriers [2, 10, 30–34]. From this one example it is clear that significant medical, public health, and cost benefits could result from PRS (e.g. [35]). It is well known that patients with atherosclerotic diseases, coronary artery disease (CAD), and lung diseases can benefit from early intervention [36–38]. ... Precision genetics is already used in identification of candidates for early intervention, and will become widespread in the near future (cf. Myriad’s riskScore test and other examples [33, 34]). In figure 4, we illustrate the predicted risk of breast cancer and coronary artery disease as function of age for high, medium and low risk groups, respectively.

We have verified in sibling studies that among two sisters the outlier with high risk score is much more likely to have breast cancer than the one with normal range score. The excerpt below is from the section on sibling validation: 
... We tested a variety of polygenic predictors using tens of thousands of genetic siblings for whom we have SNP genotypes, health status, and phenotype information in late adulthood. Siblings have typically experienced similar environments during childhood, and exhibit negligible population stratification relative to each other. Therefore, the ability to predict differences in disease risk or complex trait values between siblings is a strong test of genomic prediction in humans. We compare validation results obtained using non-sibling subjects to those obtained among siblings and find that typically most of the predictive power persists in within-family designs. Given 1 sibling with normal-range PRS score (less than 84th percentile) and 1 sibling with high PRS score (top few percentiles), the predictors identify the affected sibling about 70-90 percent of the time across a variety of disease conditions, including breast cancer, heart attack, diabetes, etc. For height, the predictor correctly identities the taller sibling roughly 80 percent of the time when the (male) height difference is 2 inches or more.
The evidence is strong that PRS outliers are at unusual absolute risk. In fact, the likelihood that an individual in the high PRS tail will eventually have the disease can approach 100% for some conditions -- see figure below. This is a concrete realization of precision medicine, at least for these individuals.

In addition to commercial products like Myriad's riskScore (which extends their BRCA panel to additional polygenic factors, and is already widely available), I am aware of many healthcare systems (including some national healthcare systems) that are seriously investigating the use of PRS in standard clinical care. 

Another example: a relative of mine had a prostate cancer diagnosis and took a (standard of care) genetic risk test which, like the pre-riskScore Myriad product, is simply a panel of rare monogenic risk variants. We and other groups have developed prostate cancer polygenic predictors which could be easily incorporated into standard of care and would likely be much more useful than the existing panel. I haven't looked carefully at the prostate cancer numbers but I strongly suspect that, as in the breast cancer example, many more men are at high risk due to high PRS than are carriers of the rare variants.

It's only a matter of time before these improvements in diagnostic screening become widespread.

Here is what we say about IVF applications:
In the past, parents with more viable embryos than they intended to use made a selection based on very little information — typically nothing more than the appearance or morphology of each blastocyst. With modern technology it has become common to genotype embryos before selection, in order to detect potential genetic issues such as trisomy 21 (Down Syndrome). Parents who are carriers of a single gene variant linked to a Mendelian condition can use genetic screening to avoid passing the risk variant on to their child. Millions of embryos are now genetically tested each year. With polygenic risk prediction, it is possible now to screen against outlier risk for many common disease conditions, not just rare single gene conditions. For example, the overwhelming majority of families with breast cancer history are not carriers of a BRCA risk variant, but rather have elevated polygenic risk. It is now possible for these families to select an embryo with average or even below average breast cancer risk if they so wish.

 Here is the paper:

From Genotype to Phenotype: polygenic prediction of complex human traits > q-bio > arXiv:2101.05870 33 pages, 7 figures, 1 table 
Timothy G. Raben, Louis Lello, Erik Widen, Stephen D.H. Hsu 
Decoding the genome confers the capability to predict characteristics of the organism (phenotype) from DNA (genotype). We describe the present status and future prospects of genomic prediction of complex traits in humans. Some highly heritable complex phenotypes such as height and other quantitative traits can already be predicted with reasonable accuracy from DNA alone. For many diseases, including important common conditions such as coronary artery disease, breast cancer, type I and II diabetes, individuals with outlier polygenic scores (e.g., top few percent) have been shown to have 5 or even 10 times higher risk than average. Several psychiatric conditions such as schizophrenia and autism also fall into this category. We discuss related topics such as the genetic architecture of complex traits, sibling validation of polygenic scores, and applications to adult health, in vitro fertilization (embryo selection), and genetic engineering.

Monday, January 18, 2021

From Genotype to Phenotype: polygenic prediction of complex human traits

New paper, prepared for the book Genomic Prediction of Complex Traits, Springer Nature series Methods in Molecular Biology.
From Genotype to Phenotype: polygenic prediction of complex human traits > q-bio > arXiv:2101.05870   33 pages, 7 figures, 1 table
Timothy G. Raben, Louis Lello, Erik Widen, Stephen D.H. Hsu 
Decoding the genome confers the capability to predict characteristics of the organism (phenotype) from DNA (genotype). We describe the present status and future prospects of genomic prediction of complex traits in humans. Some highly heritable complex phenotypes such as height and other quantitative traits can already be predicted with reasonable accuracy from DNA alone. For many diseases, including important common conditions such as coronary artery disease, breast cancer, type I and II diabetes, individuals with outlier polygenic scores (e.g., top few percent) have been shown to have 5 or even 10 times higher risk than average. Several psychiatric conditions such as schizophrenia and autism also fall into this category. We discuss related topics such as the genetic architecture of complex traits, sibling validation of polygenic scores, and applications to adult health, in vitro fertilization (embryo selection), and genetic engineering.

From the introduction:
I, on the other hand, knew nothing, except ... physics and mathematics and an ability to turn my hand to new things. — Francis Crick 
The challenge of decoding the genome has loomed large over biology since the time of Watson and Crick. Initially, decoding referred to the relationship between DNA and specific proteins or molecular mechanisms, but the ultimate goal is to deduce the relationship between DNA and phenotype — the character of the organism itself. How does Nature encode the traits of the organism in DNA? In this review we describe recent advances toward this goal, which have resulted from the application of machine learning (ML) to large genomic data sets. Genomic prediction is the real decoding of the genome: the creation of mathematical models which map genotypes to complex traits. 
It is a peculiarity of ML and artificial intelligence (AI) applied to complex systems that these methods can often “solve” a problem without explicating, in a manner that humans can absorb, the intricate mechanisms that lie intermediate between input and output. For example, AlphaGo [1] achieved superhuman mastery of an ancient game that had been under serious study for thousands of years. Yet nowhere in the resulting neural network with millions of connection strengths is there a human-comprehensible guide to Go strategy or game dynamics. Similarly, genomic prediction has produced mathematical functions which predict quantitative human traits with surprising accuracy — e.g., height, bone density, and cholesterol or lipoprotein A levels in blood (see Table 1); using typically thousands of genetic variants as input (see next section for details) — but without explicitly revealing the role of these variants in actual biochemical mechanisms. Characterizing these mechanisms — which are involved in phenomena such as bone growth, lipid metabolism, hormonal regulation, protein interactions — will be a project which takes much longer to complete. 
If recent trends persist, in particular the continued growth of large genotype | phenotype data sets, we will likely have good genomic predictors for a host of human traits within the next decade. ...

Saturday, January 16, 2021

Harvard CMSA talks (video)

I recently came across this channel on YouTube, produced by CMSA at Harvard.
The new Center for Mathematical Sciences and Applications in the Faculty of Arts and Sciences will serve as a fusion point for mathematics, statistics, physics, and related sciences. Evergrande will support new professorships, research, and core programming. 
Shing-Tung Yau, Harvard’s William Caspar Graustein Professor of Mathematics, will serve as the center’s first director. 
“The Center for Mathematical Sciences and Applications will establish applied mathematics at Harvard as a first-class, interdisciplinary field of study, relating mathematics with many other important fields,” Yau said. “The center will not only carry out the most innovative research but also train young researchers from all over the world, especially those from China. The center marks a new chapter in the development of mathematical science.”
If I'm not mistaken Evergrande is a big real estate developer in China. It's nice to see them supporting mathematics and science in the US :-) 

In 2010 I accompanied S.T. Yau and a number of other US academics and technologists to visit Alibaba, which wanted to establish a center for data science in China. Unfortunately this never really got off the ground, but CMSA looks like it is off to a good start. 

Here are some talks I found interesting. There are quite a few more.

The talk on Atiyah, Geometry, and Physics led me to this poem which I like very much. Sadly, Atiyah passed in 2019. I believe we met once at a dinner at the Society of Fellows, but I hardly knew him.
In the broad light of day mathematicians check their equations and their proofs, leaving no stone unturned in their search for rigour. 
But, at night, under the full moon, they dream, they float among the stars and wonder at the mystery of the heavens: they are inspired. 
Without dreams there is no art, no mathematics, no life. 
—Michael Atiyah

Monday, January 11, 2021

Global AI Talent Flows

The illustration above describes a global population of ~5k researchers whose papers were accepted to the leading 2019 conference in deep neural nets. To be precise they looked at ~700 authors of a randomly chosen subset of papers. There is also a more select population of individuals who gave presentations at the meeting. This is certainly not the entire field of AI, but a reasonable proxy for it.

Global AI talent tracker:
For its December 2019 conference, NeurIPS saw a record-breaking 15,920 researchers submit 6,614 papers, with a paper acceptance rate of 21.6%, making it one of the largest, most popular, and most selective AI conferences on record. 
Key Takeaways 
1. The United States has a large lead over all other countries in top-tier AI research, with nearly 60% of top-tier researchers working for American universities and companies. The US lead is built on attracting international talent, with more than two-thirds of the top-tier AI researchers working in the United States having received undergraduate degrees in other countries.   
2. China is the largest source of top-tier researchers, with 29% of these researchers having received undergraduate degrees in China. But the majority of those Chinese researchers (56%) go on to study, work, and live in the United States. 
3. Over half (53%) of all the top-tier AI researchers are immigrants or foreign nationals currently working in a different country from where they received their undergraduate degrees.
Prediction: PRC share in all 3 categories will increase in coming decades as their K12, undergraduate, and graduate schools continue to improve, and their high-tech economy grows much larger. See Ditchley Foundation meeting: World Order today

Using conference papers as the filter probably misses a lot of world class work (especially implementation at scale) that is going on in PRC at tech companies. Note in the list below the only Chinese institutions are Tsinghua and Beijing universities. But I would be surprised if those were the main accumulation of top AI talent in China, compared to large tech companies.


Saturday, January 09, 2021

Spengler (Asia Times): American Democracy died on Capitol Hill

Note although this appears in the Asia Times column Spengler, the byline is Paul Muir, not David Goldman.
American democracy died on Capitol Hill 
No Russian cyberspooks, no Chinese spies, no jihadi terrorists – no external enemies of any kind could have brought as much harm to the United States as its own self-inflicted wounds. 
I spent last evening taking calls from friends around the world, including a senior diplomat of an American ally who asked me what I thought of the first evacuation of Capitol Hill since the British invaded in 1812. “I’m horrified,” I said. “So is the entire free world,” the diplomat replied. 
There are belly-laughs in Beijing this morning. The Chinese government daily Global Times taunted: 
... The world is watching ... the country that they used to admire descend into a huge mess. Chinese observers said this is a “Waterloo to US international image,” and the US has totally lost legitimacy and qualification to interfere in other countries’ domestic affairs with the excuse of “democracy” in the future.  
[[ When protestors in HK occupied their legislature, US propaganda hailed it as a victory for democracy... when the same thing happens here it is declared domestic terrorism. ]]
It’s actually worse than the Global Times editors think. 
If it were only a matter of Trump’s misbehavior, this disaster would be survivable. The trouble is that the popular belief in a vast and nefarious conspiracy has a foundation in fact: Starting before Trump’s term in office his political opponents abused the surveillance powers of the intelligence community to concoct a black legend of Russian collusion on the part of his campaign. The mainstream media, staffed overwhelmingly by Trump’s enemies, slavishly repeated this black legend until large parts of the population refused to believe anything it read in the newspapers or saw on television. 
The leadership of the Democratic Party, its allied media, and the Bush-Romney wing of the Republican Party decided to play dirty to expunge an obstreperous, incalculable outsider from the political system. And in doing so, this combination, America’s establishment, destroyed public trust in the Congress and the media. It’s no surprise that two out of five Americans now believe that a vast conspiracy rigged the 2020 presidential elections. 
The spectacle of a serving president inciting a mob against the US Congress to stop the certification of his successor held the world in morbid fascination. But the biggest problem isn’t Trump’s misbehavior, egregious as it is, but the eruption of popular rancor against the constitutional system that has made America a model of governance for the world. Leftist mobs last spring burned police stations and destroyed shopping districts in a rampage against supposed systemic racism, and Trump supporters desecrated the Holy of Holies of American democracy, the chamber of the United States Senate. 
Behind the minority of violent actors is a majority that believes the system is rigged against them – whoever “them” might be. The Democrats say that the system is rigged against African-Americans, women, and other minorities, and the Republicans say that a global elite has rigged the system against middle-income Americans. “Rigged elections” has the same resonance as “systemic racism.” These by-words imply that disagreement is prima facie proof of villainy: To deny that there is systemic racism is to be a racist, and to deny that elections are rigged is evidence of complicity in a vast plot. 
A quarter of Americans believe that Covid-19 was a planned conspiracy of one kind or another, according to the Pew Survey; just under half of Americans with a high school education or less believe this. One out of three believes that a “deep state” is trying to undermine Trump. I reject the first and believe the second: my colleagues at Asia Times and I have regular access to virologists in a number of countries with scientific credentials and no political agenda to pursue, and can sift scientific evidence and opinion. By contrast, I know personally enough of the actors in the so-called “deep state” to conclude that they are acting in concert to wreck the Trump Administration. I also know many of the writers who have exposed the “deep state,” including Andrew McCarthy and Lee Smith, to trust their bona fides. I denounced this conspiracy repeatedly in these pages, most recently in an essay entitled “The Treason of the Spooks” (Dec. 4, 2020). For details, see Andrew McCarthy’s 2019 book Ball of Collusion, which I reviewed in Asia Times, or Lee Smith’s The Plot against the President. 
Sometimes there is a conspiracy and sometimes there isn’t. But Trump’s political supporters, bombarded daily by fake news about Russian collusion and other alleged misbehavior, have come to distrust any criticism of their president. 
If Trump was right that the whole impeachment business was an extra-legal conspiracy on the part of his enemies, why shouldn’t they believe that the election was rigged? This is a lose-lose proposition. Assume that Trump is right, and the election was rigged. In that case the United States has become a banana republic and American democracy a twisted joke. Assume that he is wrong, and that nonetheless – as Sen. Ted Cruz (R-Texas) intoned to justify his refusal to accept the election outcome – 39% of Americans nonetheless believe that election has rigged, because their president told them it was rigged. In that case the public trust that makes democracy possible has collapsed. The people, as Bertolt Brecht observed after demonstrations against the Soviet puppet government in East Germany, have lost the confidence of the government, and the simplest course of action would be for the government to dissolve and for the people to elect a new one. 
... Americans are frightened for their future, with good reason. They see enormous rewards accrue to a handful of tech companies, and stagnation and decay in large parts of the rest of the country. Donald Trump gave them a frisson of hope, and the Establishment reaction against Trump confirms the popular suspicion that a malevolent global elite has seized control of their country. Trump shamefully exploited this suspicion to direct a popular storm against the Congress. 
The US is living off borrowing from the rest of the world. Its net international investment position fell by about $12 trillion during the past 10 years. And the federal deficit is now 15% of gross domestic product, the highest since World War II. What can’t go on forever, won’t (in the late Herb Stein’s famous formulation).

Thursday, January 07, 2021

YouGov on storming of capitol

I misread the last line of this when I first saw it... I thought 56% of all voters polled believed the election was stolen. Probably what it actually says is that 56% of people who believe the election was stolen think storming the capitol is OK.

These results suggest 30-40% of all voters think the election was stolen -- i.e., roughly 2x the number who approve of the storming:  0.21 / 0.56 ~ 38%

YouGov poll of 1,397 registered voters.

Wednesday, January 06, 2021

The Last Emperor

1987 seems so long ago. Watch this movie! 

Nine Academy Awards, including Best Picture, Best Director, and Best Score.

See also 

Twilight in the Forbidden City  (account of Sir Reginald Fleming Johnston, tutor to the last emperor of China)

Sunday, January 03, 2021

Two from Spengler (David Goldman at AsiaTimes)

David Goldman, a former banker, writes the Spengler column for AsiaTimes, where he is business editor. 

Huawei 5G in Germany, Japan, and S. Korea? 

Book Review: American Awakening: Identity Politics and Other Afflictions of Our Time, by Joshua Mitchell (Georgetown University) 

2. AsiaTimes tells Le Figaro why China is winning the tech war (interview)
LM: Germany just announced that it will allow Huawei 5G to be installed. What conclusions do you draw from this decision? Is this short-term logic, that will hand the control of big data to China? 
DG: To my knowledge, Germany has made no announcement, but the German media have leaked the draft law that the government will present to the Bundestag, which allows Huawei 5G. Trump’s defeat in the US election probably tipped the balance in favor of Huawei. Huawei always has viewed 5G as the core of an “ecosystem” of new technologies that 5G makes possible. ... 
LM: Obama had launched the Trans-Pacific Partnership. Now there is a China-led trade zone, the RCEP. Have Australians, South Koreans and others decided to go back to China in a realpolitik move, because they see America as a declining power, engulfed in internal wars and not to be trusted? 
DG: The Regional Comprehensive Economic Partnership will cut tariffs dramatically – by about 90% in the case of Japanese exports to China – and now China is trying to negotiate free trade areas with South Korea and Japan. Asian trade is now as concentrated within Asia as European trade is concentrated within Europe. 
The logic of the development of an Asian internal market is similar to that of the European Community, and it is not surprising that the Asians are creating a giant free trade zone. Australia is in a nasty fight with China, but it now sells a higher proportion of its exports to China than ever before. It could not afford to stay out of the RCEP. 
The American consumer for decades was the main source of demand in the world economy. Now the internal Asian market is far more important. South Korea, for example, exports twice as much to China as to the US. I am sure that the Japanese and South Koreans like the United States much better than they like China, but the economic logic behind an Asian free trade zone is overwhelming. 
An Asian free trade zone certainly is compatible with America’s role as the leading superpower, just as the European Community originally was formed with American sponsorship during the Cold War. 
The difference, of course, is that China’s economic strength makes it a magnet for all the Asian economies. In this context, it is noteworthy that Japan and South Korea politely rejected American demands to exclude Huawei from their 5G networks. 
To restore high-tech manufacturing, we would need the sort of tax credits and subsidies for capital-intensive industry that Asian governments provide; we would need the sort of support from the Defense Department that led to every important technology of the digital age, from microprocessors to the Internet; and we would need a greater emphasis on mathematics and science at every level of education. 
Above all, we would need the sense of national purpose that John Kennedy evoked with the space program or Reagan with the Strategic Defense Initiative. Considering that we have just spent several trillion dollars subsidizing incomes and supporting capital markets, another trillion dollars to support technological superiority doesn’t seem extravagant. ...

Blog Archive