This yr, I’ve a really lengthy commute. As soon as per week, I must drive about 2 hours to succeed in a shopper’s workplace, then drive again 2 hours. It sounds horrible, however I find it irresistible as a result of throughout these 4 hours, I can hearken to all of the podcasts I like in peace. No one complaining, nothing else to do, aside from listening to geeky science stuff. That is the life 🙂
Right here is the record of what I’ve been listening to not too long ago. It is a good record for bioinformaticians seeking to be taught extra about machine studying, in addition to ML professionals taken with bioinformatics. There’s a mixture of technical and extra normal facets associated to drug discovery.
Cátia Moutinho’s podcast about Single-Cell is a incredible useful resource for staying up-to-date with the advances on this subject. This subject has been rising so quick, with many new platforms and instruments being launched in the previous couple of years, that it is rather tough to comply with what’s occurring. This podcast has been a useful useful resource for me, saving me hours of analysis time. There are introductory episodes, the place Catia goes via the fundamentals of spatial or reviews the different platforms, to extra specialised episodes the place she interviews a visitor and goes into particulars of a selected know-how. The host comes from a wet-lab background, which, for me, as a bioinformatician, may be very helpful, because it helps me perceive the opposite aspect. Sadly, this podcast has been not too long ago discontinued by the host. If you happen to’re studying this, Catia, thanks for the episodes — and I actually miss the podcast! 🙂
Nearly all the things in drug discovery is expounded to immunology. But, immunology is such an enormous subject that it’s obscure all of it. I may spend my complete life finding out the immune system, and I nonetheless wouldn’t know all the things. This podcast is an incredible useful resource to be taught and sustain with the latest advances. In each episode, the hosts evaluation 4 new papers, then they invite a particular visitor. One of many hosts may be very keen on CD4+ cells, with good causes, however they don’t overlook about controversial macrophages, mischievous neutrophils, smart B cells, and all the opposite cell sorts in our immune system. My favorite episode is the 2 hours special with Dr Peter Doherty, Nobel Prize winner, speaking concerning the discovery of the position of killer T-cells, finished earlier than articles had been digital.
It is a top-class report about drug discovery and the pharma business, by journalist Daniel Levin. You could hearken to this if you wish to see the large image about which medication are being developed and what are the newest traits. My favorite episode is this interview with Jorge Goldstein, concerning the historical past of authorized battles in drug discovery — reminiscent of when genes could possibly be patented, till the ruling was reversed in 2013.
I believe that each scientist ought to discover the time to hearken to the Nature Podcast, or a minimum of learn the information and feedback part on the Nature web site. That is an incredible Podcast, not just for maintaining with latest publications, but in addition for understanding the politics of science total. That is my method to hearken to what’s going on within the US with the latest unlucky political information, to listen to the influence of those from individuals affected by them.
Longevity is an enchanting subject. It’s a distinct strategy to drug discovery: as an alternative of specializing in a single illness, you look at a mixture of circumstances, together with comorbidity, and the way varied ailments and damaging elements work together with one another. There are numerous podcasts about longevity, however a lot of them are poor high quality, scientifically talking. This one, from Hannah Went, is without doubt one of the few I contemplate good high quality, particularly in case you are within the evaluation of methylation. I’ve discovered quite a bit from it, and it has saved me time.
My background is in Genetics, and this podcast from Sano Genetics, an organization growing therapies for uncommon ailments, is one among my favourites. If you wish to find out about GWAS and their purposes, it is a should. One other good podcast is the Illumina Podcast, which isn’t revealed very ceaselessly, however options good interviews on the purposes of sequencing for drug discovery.
The Chain — Protein Engineering Podcast
I’ve began listening to this podcast not too long ago, as I wished to be taught extra about protein design. It is a fascinating subject, which can be bioinformatics, but I have no idea a lot about it. My favorite episode is this special about using AI/ML for antibody design, which explains that AlphaFold 3 just isn’t sufficient to unravel the issue.
I’ve reached the conclusion that one of the simplest ways to find out about machine studying is to hearken to the individuals who develop AutoML instruments. The individuals who develop AutoML options have nice expertise in growing fashions, and so they know all about cross-validation, bagging, stacking, hyper-parameter-optimization, function engineering, and all the things wanted to make sure your mannequin is skilled correctly. Even if you happen to don’t plan on utilizing AutoML instruments, there’s a lot to be taught from listening to those conversations. My absolute favorite episode is the 4 hours particular on the history of AutoGluon — it took me two weeks to complete it listening to it, however oh expensive I used to be so completely happy once I did. I additionally favored this one on TabPFN — it was good to hearken to the story of the creator, from H2O to SparkML, to basis fashions (however TabPFN didn’t work properly once I tried it). There may be additionally this one on Foundation Models, from the one that coined the time period. And this different one on Neural Network Architecture Search. I ought to cease earlier than I find yourself writing about all of the episodes :-).
DataFramed is a strong podcast, which I’ve been listening to for years. It’s in all probability the very best useful resource for following what is going on on within the LLM world. That is the place I heard about chatGPT for the primary time. However there are additionally many good episodes on information, on Vector Databases, Data Graphs, and infrastructure. There are additionally glorious episodes on coverage, such because the one on the EU act, and profession growth.
One other related podcast collection that has been in my record for years is Knowledge Skeptic. This focuses extra on particular purposes, and deeps down into the technical particulars of a ML downside.
The Effective Statistician and More or Less
I really like listening to statisticians. More or Less from the BBC is completely one among my favourites. The hosts choose up statistics cited by politicians or newspapers, and so they confirm the numbers. It’s a chunk of wonderful journalism. It’s enjoyable listening to them debunking all of Trump’s disparate numbers. I want I had a podcast like this in Italy, once I was youthful.
One other Statistics-related podcast is The Efficient Statistician, which ceaselessly focuses on using statistics for drug discovery. One in all my favorite episodes is this one concerning the position of statisticians in the course of the medical trial submission course of — I didn’t know that the method was so strict and demanding, and so susceptible to misunderstandings. One other favorite episode is this one about P-values — I really like listening to statisticians complaining concerning the misuse of P-values :-).
