laserson

I read this last year and also recommend!

Nice essay Ankit!

I think there are two more factors worth touching on that make things difficult in drug discovery:

1. It feels like you're differentiating a bit but I always hear of the APIs being leaky. The selection of a target in the target discovery phase may drive which modalities are even available to try to engineer an intervention (e.g., intracellular vs extracellular targets can be addressed with different types of drugs). Clinical factors can also determine whether a particular modality is appropriate (e.g., keeping drugs refrigerated).

2. Costs create constraints that need to be applied at all the stages of the pipeline. Does gathering evidence during target discovery require a unique assay or access to very rare patient samples? What is the size of the addressable market? If you can only target a narrow patient population, it may not end up justifying the costs involved in running the clinical trials. If your chosen drug modality costs $500k to synthesize one dose, how will it be paid for?

There’s a well-known Louis Pasteur quote that I always liked:

“Chance favors the prepared mind”

I'm a big fan of Tiller [1], which is a paid service that basically just connects Yodlee to Google Docs or Excel. It will dump all your transactions and balance info into a spreadsheet and you can do what you want with it. It also has lots of templates for things like budgeting etc. I run a daily cron job that loads the transaction data and produces a nice Jupyter-based report. It will require some custom building, but the API plumbing is done for you.

[1]: https://www.tillerhq.com/

One thing I miss from Google is relatively seamless integration with maps. Sometimes the address queries just fail for me on Kagi. And even when they don’t, Google’s maps are still far more useful IMO.

In Israel they produce single-serving chocolate milk bags, which now mainly serve as a fond dose of nostalgia for me.

I’ve found Citymapper to give far more realistic routes and commute times on the NYC subway.

How does this compare with Concertmaster? It's basically a skin on top of Spotify for browsing classical music. It's free and pretty good. (There's also an app.) And I believe there is a version for Apple Music already?

[1]: https://concertmaster.app

Patch Bio | Computational Biologist | New York City | Full-time | REMOTE

Patch Bio engineers DNA for gene therapy. We build technology to make gene therapies safer and more effective, and able to address a broader range of diseases. To do that, we work at the interface of genomics, machine learning, and DNA synthesis. We are a seed-stage, venture-backed biotech startup based in New York City.

We're looking for a Computational Biologist who will be the connective tissue between the wet and dry lab. You will design DNA libraries, analyze sequencing data, and build predictive models of genetic regulation.

Check out a more detailed job description here (with contact info): https://bit.ly/patch-bio-comput-biol

Hadn't heard of Primephonic, but I have been using Concertmaster.app as a classical music-focused "skin" for Spotify and it's pretty good (and free!).

I would love to see how this was implemented. Is this on GitHub or equivalent somewhere?

It wasn't too fancy. I bought self-adhesive laminating sheets and reinforced with "book tape".

https://photos.app.goo.gl/QRCgTCUas1uYnY6x5

It did feel a bit sacrilegious slicing through the binding, but overall it worked out great!

I used an X-Acto knife to slice it into 3 volumes and rebound them.

Altair does Grammar of Graphics and interactivity, and has ~6k stars.

I always thought Go might be a good candidate for some of these tasks.

Alchemab Therapeutics | London, Cambridge UK | Software/Data Engineer | ONSITE, VISA | Full-time

Alchemab Therapeutics is a VC-backed newco that is bringing together the latest advances in antibody repertoire profiling, deep learning, and synthetic DNA libraries to discover new antibody-based drugs in a variety of disease areas including oncology and neuroscience. Alchemab aims to deliver a pipeline of targets and therapeutic candidates over the next 18 months.

We are looking for a data engineer with cloud expertise to help build and maintain our cloud/software infrastructure. This person will work closely with our bioinformatics/statistics experts to build reliable data processing pipelines and define/implement software engineering/devops best practices.

Experience with at least some of the following is a requirement:

- AWS stack, including administration and various database products

- Data storage, (relational) data modeling, workflow engines and schedulers

- Python data stack

- Distributed computing engines, such as Spark or Dask

- Docker containerization

Experience with bioinformatics, next-generation DNA sequencing (NGS), and genomics pipelines is a plus.

The ideal candidate will feel comfortable taking the lead at a small startup to define and implement devops and data engineering tasks.

If you're interested, send an email to uri@alchemab.com.

Alchemab Therapeutics | London, Cambridge UK | Software/Data Engineer | ONSITE, VISA | Full-time

Alchemab Therapeutics is a VC-backed newco that is bringing together the latest advances in antibody repertoire profiling, deep learning, and synthetic DNA libraries to discover new antibody-based drugs in a variety of disease areas including oncology and neuroscience. Alchemab aims to deliver a pipeline of targets and therapeutic candidates over the next 18 months.

We are looking for a data engineer with cloud expertise to help build and maintain our cloud/software infrastructure. This person will work closely with our bioinformatics/statistics experts to build reliable data processing pipelines and define/implement software engineering/devops best practices.

Experience with at least some of the following resources is a plus:

- AWS stack, including administration and various database products

- Data storage, (relational) data modeling, workflow engines and schedulers

- Python data stack

- Distributed computing engines, such as Spark or Dask

- Docker containerization

The ideal candidate will feel comfortable taking the lead at a small startup to define and implement devops and data engineering tasks.

If you're interested, send an email to Olivia, our head of operations: olivia@alchemab.com.

I like TillerHQ (https://tillerhq.com/) for just extracting the financial data and dumping into a spreadsheet.

Also, the PCR reaction was effectively described in an earlier 1971 paper from Gobind Khorana (see final paragraph): https://www.sciencedirect.com/science/article/pii/0022283671...

More color on Kary Mullis: https://classic.esquire.com/article/1994/7/1/is-kary-mullis-...

Strang just published a new book called "Linear Algebra and Learning From Data" [1] which I only just started but find to be quite enjoyable so far. It's simultaneously conversational but also quite terse (similar to "All of Statistics"). There are many advanced and very contemporary applications covered in the book, with a focus towards machine learning.

[1]: https://www.amazon.com/Linear-Algebra-Learning-Gilbert-Stran...

Strongly agree. This was actually the text book when I took numerical methods at MIT. As far as math texts go, it is definitely a joy to read.

Hadley Wickham

So many people use Excel/Google Sheets as a database rather than a spreadsheet, except you don't get any of the advantages of a database. It's astounding to me that there hasn't really been a webapp that's essentially just a nice collaborative frontend for a SQL database, which is how I use Airtable in my lab. I'm very happy to see they're on a sustainable path, and I hope they continue to support the use case of just-a-database. Congrats!

Congrats! We've had a good amount of fun playing with the OT-One (and also some frustrations). I think the large range of volumes is great, along with what seems to be more streamlined/accurate pipetting. I think the enclosure is also a solid feature.

One of my favorite features is the ability to use a simple Python API to define a protocol. Will this still be the case or do people have to use the GUI?

Is there a reasonable (cheap) path to upgrade?

Is it possible to have a notebook and a console using a single, shared kernel?

>I also thought it was funny that she complained about the committee not raising those questions when she submitted her proposal years ago.

At least in the sciences, your thesis committee is supposed to continually/regularly keep tabs on your progress, and help ensure you're making progress towards a passing dissertation. This is the committee's responsibility.

>An approved proposal does not in any way imply a seal of approval for every detail of the dissertation... come on now.

For sure, but the committee should know how things are progressing. In the sciences, the thesis defense is not really a "defense". It only gets scheduled when everyone basically agrees you will pass, and it's more of a celebration. It's very rare to have hostile defenses, and extremely rare for someone to fail. (If that happens, it's usually due to politics.)

>As if that is a valid response to this sort of response?

True, but it is concerning if they're bringing up possible plagiarism by her and Dr. Mao, while not following through on any kind of punishment/censure of the two of them.

The Hail project at the Broad Institute makes use of Kudu I believe.

https://hail.is

Not sure if that group is paying for CDH, but this tool is definitely being built for others to use.

And separate from the issue of whether it's a good or bad idea, it will shortly be possible to send raw SQL queries to the backend.

https://github.com/cloudera/ibis/issues/412