Lior Pachter, a computational biology and math professor at Caltech, critiques statistics and methods in genomics, bioinformatics, and published research.
HN commenters cite it for detailed analyses of dubious studies, like a vitamin D trial and network science claims, and for explaining misleading genetic test results.
That 40% number is accurate but misleading. Here's an old prediction of pretty much that result that explains the issue:
https://liorpachter.wordpress.com/2013/11/30/23andme-genotypes-are-all-wrong/
Basically, almost all (~99.9%) of the results of tests like 23andme (SNP arrays) are correct, but because an interesting/scary variant/mutation is about as rare in the population as the test result being wrong for a given site, somewhere around half of those scary results are false positive. If 1000 people are tested for a mutation with this method, 10 will be false positive. The real frequency of mutations vary by site - some will be close to 50%, others at 10e-9, depending on how strongly that variant is selected against, so the most harmful ones are very rare. If you get a result from 23andme that your doctor thinks is worth retesting, it's probably because it's a harmful mutation, which is rare, and ergo there's a fair chance the 23andme result is wrong. So 23andme is almost always right, except when the result might affect your health in a very serious way. I should point out that for late-onset diseases like cancer, heart disease, alzheimer's, etc., there's little natural selection, so pathogenic variants are more common, making it more likely that the 23andme result is correct.
It's a simple bayes rule result. It would not be hard for 23andme to give you an estimate of how likely the result is to be a false positive, but maybe there's some regulatory barrier in the way?
It's tricky to pinpoint exactly when modern "bioinformatics" started. Some notable times are 1979 when the Los Alamos Sequence Database started, 1982 when that became GenBank, and certainly by 1990 when the Human Genome Project started.
GenBank is "an open access, annotated collection of all publicly available nucleotide sequences and their protein translations" (https://en.wikipedia.org/wiki/GenBank ). Most journals now require that sequences be entered into GenBank before publication.
I'll add a few historical observations. The early work at Los Alamos was on Sun machines running Sybase (1987, according to https://en.wikipedia.org/wiki/Sybase ). These are serious Unix fans, who quickly took to perl when it came out. I think some of the support for SybPerl came from bioinformatics; the SybPerl, OraPerl, and other *Perl systems helped create the extension system for perl5. (I am not able to verify that, though the author of SybPerl, Michael Peppler, consulted in the mid-1990s for "Research Genetics" http://www.peppler.org/resume.html .)
Unix people in the early 1990s were serious perl fans. The most popular perl4 library for CGI programming was cgi-lib.pl, by Steven Brenner, a computational biologist/bioinformatics researcher. It was replaced in perl5 by CGI.pm, by Lincoln Stein, another bioinformatics researcher.
This gives a hint that bioinformatics has not only a close connection to the technologies needed for the first dot-com era, but also that the field itself…
(Somewhat OT, but...)
Constructive logic accepts the truth of every integer is either even or odd, but only because we have
an effective means of determining which alternative holds for any integer. It does not accept the statement every real number is either rational or irrational; given a real number, say as a convergent series, we have no effective means of determining whether or not it is rational.
I don't understand this. Consider the smallest number N so that any N points in the coordinate plane in general position contain a subset of 7 points that form the vertices of a convex polygon (taken from [0], recently discussed here on HN): is this number, which is definitely an integer, even or odd? We don't know, because we don't know what the number is. Isn't that the same situation we are in with regard to the convergent series? What's the difference? If we actually computed the decimal expansion of the value of the convergent series, we would be able to look at it and see whether it was rational or irrational. There is the minor difficulty that the expansion is in most cases going to be infinite, but so what? We're talking about mathematics here, not computation.
[0] https://liorpachter.wordpress.com/2015/09/20/unsolved-problems-with-the-common-core/, "Fifth grade" section
>Failure to distribute the data and code not only greatly devalues the contributions of many scientists, it makes replication far more difficult, and opens the door for outright fabrication.
Reasons why labs don't produce code:
(1) Fear of being scooped; suppose you put in tons of man-hours into a custom population genetics association study for malaria in Africa, you don't want your competitor lab to sequence a bunch samples in SE Asia and run your code as-is and publish a paper when you have the sequencers to do the same.
(2) Fear of competitors not being able to replicate stochastic ML results; some machine-learning and artificial neural network papers applied to Biology are stochastic. Past authors have been accused of "cherry-picking" the most "optimistic" runs that show positive results, e.g., (https://liorpachter.wordpress.com/2014/02/11/the-network-nonsense-of-manolis-kellis/)
(3) Focus on science, not tool production; most labs' focus is to produce publications, not open-source software. User base for scientific software is very niche, unlike Web MVC frameworks; making the payoff calculus of packaging and supporting external users not so great. Furthermore, most labs have custom databases making integrating external software, externalizing internal software difficult.
It's like finance or law or any other field with some hothead cutthroat personalities in powerful positions. Where there is perceived power there is an attraction to people who feed off of power alone. Egos are a plenty and often cruel to others trying for the same goals. Lots of people are hypercompetitive rather than collaborative, and see any peer as an adversary or challenger rather than leaning into synergy.
This kind of stuff would happen at conferences too, a couple notable cases of screaming matches from the audience to the presenter have happened. Sometimes vindictive professors will go to a rival professors's grad student's talk and just grill the poor kid, asking them impossible questions and being hypercritical. Usually the grad student's own professor has to step in and try to deflect the assholes line of questioning. Sometimes the professors are even in the same department. It's a total cringe moment for the rest of the audience, and is definitely a phenomenon that is slowly growing out of the field as professors retire.
Take a look at the comment section of this blog to see it first hand: https://liorpachter.wordpress.com/2017/08/02/how-not-to-perform-a-differential-expression-analysis-or-science/
Completely agree with your comment, and here you have some more examples in the Spanish university to support your thesis:
- Rafael Luque was suspended of employment for 13 years not because of blatant dishonesty but because he was signing with the King Abdulaziz University as first affiliation and that bothered his university [1]. The KAU is known for buying citations this way [2].
- Another example is Francisco Herrera Triguero [3]. He's been signing with KAU as second affiliation for years and nothing happened to him even when the story about the other guy broke out.
So yeah, maybe we don't need more algorithms but more public oversight over this stuff.
[1] (Spanish): https://elpais.com/ciencia/2023-03-31/suspendido-de-empleo-y-sueldo-por-13-anos-uno-de-los-cientificos-mas-citados-del-mundo-el-espanol-rafael-luque.html
[2] https://liorpachter.wordpress.com/2014/10/31/to-some-a-citation-is-worth-3-per-year/
[3] (Spanish): https://www.elsaltodiario.com/universidad/papers-y-mas-papers-las-sombras-en-la-industria-de-las-publicaciones-cientificas