You don't have to learn about data science methods from a textbook but we think it is an effective approach. There are five main ways we learn about data science techniques:
Google is great if you want to know something specific and you already know the name of that specific thing, but it is less good for building up your general knowledge of related subjects. Papers can be effective, but they often lack necessary background, tend to be very specific, and are often written by writers whose first priority is not to provide a clear explanation. Colleagues are rarely authoritative on a particular subject, and may not be interested in discussing their knowledge. Finally, lectures are great but it is hard for them to be paced appropriately for everyone and often do not come with the necessary exercises to develop technical skills. Hence we prefer the traditional textbook, which mostly just requires a disciplined student.
It's hard to be exhaustive when it comes to a subject as broad as data science. If you have topic suggestion, let us know here. That being said, please keep your arguments that algebraic topology is a core data science subject in your NSF grant application, not our e-mail.