Episode 287 Alan Rubin on MaveDB - CourtneyR.dev's 2nd Brain

Episode 287: Alan Rubin on MaveDB ![rw-book-cover](https://img.youtube.com/vi/tEFJ9yqsCBw/maxresdefault.jpg) 🔗 **Source:** [Episode 287: Alan Rubin on MaveDB](https://www.youtube.com/watch?v=tEFJ9yqsCBw) > [!note]+ Document Notes > tags: #computational-biology #open-science #research-software #medical-research #cancer-research #data-infrastructure #django #python #clinical-genetics #sustainability > author: Sustain Podcast > source: YouTube > reference: https://www.youtube.com/watch?v=tEFJ9yqsCBw > date created: 2024-12-19 15:30 > > # 🏷️ Title: Episode 287: Alan Rubin on MaveDB >  ## 🧰 Action Inbox - [H1] Explore MaveDB for protein variant data — research database for medical/biological researchers #computational-biology - [H2] Consider PyconAU for scientific computing skills — cross-domain learning opportunity #python - [H3] Review G4GH data standards for genomics — international health data interoperability #open-science ## 🧭 Table of Contents - [[#📝 Summary]] - [[#📖 Main Content]] - [[#🗒️ Notes]] - [[#📚 References]] - [[#🗣️ Paraphrase]] - [[#📰 Newsletter]] - [[#✅ Action Items]] - [[#🧪 Hypotheses]] - [[#📊 Methodology]] - [[#❓ Questions]] - [[#🔮 Future Research]] - [[#💡 Implications]] - [[#🧠 Learning Styles Assessment]] - [[#📊 Accessibility Assessment]] - [[#🌍 DEIB Assessment]] - [[#✅ Logical Validity]] - [[#🛡️ DARVO Analysis]] - [[#🎯 Conclusion]] ### 🔗 References - [MaveDB Database](https://mavedb.org) - [Atlas of Variant Effects Alliance](https://varianteffect.org) - [Global Alliance for Genomics and Health](https://ga4gh.org) ## 📝 Summary - Alan Rubin, computational biologist at University of Melbourne, discusses MaveDB - an open database for storing protein variant effect data - Project started 10 years ago to solve data discoverability problem in multiplexed assays of variant effects (MAVE) field - Built using Django after inspiration from PyconAU conference, serves both researchers and clinicians - Faces classic infrastructure funding challenges - everyone wants it to exist but few want to pay for it - Recently expanded with MaveMD interface specifically for clinical geneticists to look up individual variants - Bottom line: Research software sustainability requires community building, cross-domain collaboration, and creative funding strategies ## 📖 Main Content MaveDB addresses a critical gap in biological research data infrastructure. When scientists perform multiplexed assays of variant effects (essentially "fuzz testing" proteins by testing thousands of genetic variants simultaneously), the resulting data was historically scattered across supplemental tables attached to research papers, making it nearly impossible to discover and reuse. Alan Rubin and collaborators built MaveDB as a centralized repository using open-source tools, particularly Django. The project exemplifies how research software can bridge domains - serving both basic researchers studying protein function and clinical geneticists trying to interpret patient genetic variants. The sustainability challenge is classic infrastructure funding: high utility but difficult to monetize or fit into traditional research grant categories. The team has succeeded through collaborative funding models, embedding database support into larger research grants, and building an essential position in the data supply chain. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/tEFJ9yqsCBw?t=0) — [H1] Introduction to MaveDB project - Key idea: Database for storing multiplexed assays of variant effects data - Evidence: "Research database for medical researchers, biological researchers, biochemists, people interested in molecular biology and protein science" #### ⏱️ [03:45](https://youtu.be/tEFJ9yqsCBw?t=225) — [H2] Protein function and genetic variants - Key idea: Proteins are cellular machines whose function depends on amino acid sequence determined by DNA - Evidence: "All activities being carried out in your body are mediated by proteins... function is determined by its sequence" #### ⏱️ [06:30](https://youtu.be/tEFJ9yqsCBw?t=390) — [H2] Combinatorial testing challenge - Key idea: Testing protein variants creates massive combinatorial problem requiring high-throughput methods - Evidence: "For each amino acid you've got 20 different options at each position. So now you've got a 20N combinatorial problem" #### ⏱️ [08:15](https://youtu.be/tEFJ9yqsCBw?t=495) — [H1] Data discoverability problem - Key idea: Variant effect data was scattered across supplemental tables in different formats - Evidence: "These were very difficult to discover. They were all in very different formats and it was kind of getting out of control" #### ⏱️ [11:20](https://youtu.be/tEFJ9yqsCBw?t=680) — [H1] Django inspiration from PyconAU - Key idea: Regional conference attendance directly enabled project creation - Evidence: "I went to PyconAU... saw a talk from Russell Keith McGee about Django... I told Doug that I could solve this problem" #### ⏱️ [13:45](https://youtu.be/tEFJ9yqsCBw?t=825) — [H2] Django development success - Key idea: Following Django tutorial enabled rapid prototype development for scientific database - Evidence: "We quite literally followed the Django girls tutorial, but instead of building a blog, we built a database for research scientists" #### ⏱️ [16:30](https://youtu.be/tEFJ9yqsCBw?t=990) — [H1] Infrastructure funding challenges - Key idea: Infrastructure projects face systematic funding difficulties despite clear utility - Evidence: "Infrastructure is something that everybody wants to have exist and nobody really wants to pay for it" #### ⏱️ [19:45](https://youtu.be/tEFJ9yqsCBw?t=1185) — [H2] Limited funding opportunities - Key idea: Few funding mechanisms specifically target software infrastructure in life sciences - Evidence: "There are relatively few funding opportunities directly for infrastructure and for software infrastructure" #### ⏱️ [23:10](https://youtu.be/tEFJ9yqsCBw?t=1390) — [H1] Clinical genetics pivot - Key idea: Clinical community demanded different interface for individual variant lookup - Evidence: "They wanted to know, tell me everything you know about this specific DNA change. That's how they wanted to experience the data" #### ⏱️ [28:30](https://youtu.be/tEFJ9yqsCBw?t=1710) — [H2] Role transition to community building - Key idea: Project leadership evolved from technical to community-focused work - Evidence: "My role has transitioned from being more technical in the early days to being almost entirely community focused" #### ⏱️ [32:15](https://youtu.be/tEFJ9yqsCBw?t=1935) — [H2] Team structure and sustainability - Key idea: Distributed team with defined roles enables sustainable development - Evidence: "We have a team of four people... senior developer whose primary job it is to work on MaveDB" #### ⏱️ [36:20](https://youtu.be/tEFJ9yqsCBw?t=2180) — [H3] Community contribution challenges - Key idea: Difficult to engage hobbyist contributors in specialized research software - Evidence: "We don't really get community contributions... it's just not that kind of project" #### ⏱️ [42:10](https://youtu.be/tEFJ9yqsCBw?t=2530) — [H1] Value of cross-domain conferences - Key idea: Scientists benefit from attending software engineering conferences to improve craft - Evidence: "If I go to a biology conference... I'm not going to learn anything about how to do what I do on a day-to-day basis" #### ⏱️ [47:30](https://youtu.be/tEFJ9yqsCBw?t=2850) — [H2] Open source enabling academic research - Key idea: Academic projects depend entirely on free, well-documented open source tools - Evidence: "For every large company website... there are hundreds of other little academic projects that only exist because all of this stuff is freely available" ## 📚 References - [MaveDB Platform](https://mavedb.org) - [MaveMD Clinical Interface](https://mavedb.org) - [Atlas of Variant Effects Alliance](https://varianteffect.org) - [G4GH Data Standards](https://ga4gh.org) - [Sustain Podcast](https://sustainoss.org) ## 🗣️ Paraphrase - MaveDB solves a data fragmentation problem by centralizing protein variant effect measurements that were previously scattered across research paper supplements - The project demonstrates how attending cross-domain conferences (like PyconAU) can directly enable research breakthroughs by exposing scientists to appropriate technical tools - Research software sustainability requires building essential infrastructure that becomes difficult to replace, then leveraging that position for collaborative funding ## 📰 Newsletter 🧬 **Research Meets Open Source: The MaveDB Story** Ever wondered how open source tools help fight cancer? Alan Rubin's MaveDB project shows the powerful intersection of Python web development and life sciences research. Starting from a Django tutorial attended at PyconAU, the team built a critical database that serves both researchers studying protein function and doctors interpreting patient genetic variants. **Key Takeaway**: Cross-domain learning pays dividends. Scientists attending developer conferences gain technical skills that transform their research capabilities, while developers can contribute to world-changing health applications. **Practical Applications**: - Research projects can leverage mature web frameworks like Django/FastAPI for rapid prototyping - Regional conferences provide valuable cross-pollination opportunities - Infrastructure sustainability requires community building and creative funding approaches Resources: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### DevRel - Developer advocacy: Showcase research software use cases at Python conferences, create "science track" content - Education: Develop tutorials bridging web development and scientific applications - Developer experience: Document pathways for contributing to research software projects - Community management: Facilitate connections between hobbyist developers and research teams #### OSPO - Community Engagement: Support academic open source through institutional memberships, conference sponsorships - Strategic Alignment: Track research software in organizational SBOM, measure health/science impact - Developer Support: Provide consultation for academic teams adopting enterprise development practices #### Community - Contributors: Create "good first issue" pathways for research software projects - Ecosystem: Support cross-domain conferences that enable knowledge transfer between fields #### WordPress - Community: Explore health/science applications of WordPress in research data publishing - Documentation: Document WordPress use cases in academic and research contexts ## 🧪 Hypotheses - Cross-domain conference attendance accelerates innovation by exposing specialists to tools from adjacent fields - Research software sustainability improves when projects become essential infrastructure rather than seeking direct monetization - Clinical translation of research tools requires fundamentally different user interfaces and mental models than research applications - Academic open source projects struggle with hobbyist contributions due to domain complexity and mission focus ### 📊 Methodology - Survey PyconAU attendees about cross-domain applications of learned techniques - Analyze funding patterns for research infrastructure vs discovery-focused grants - Compare user interface requirements between research and clinical genetics workflows - Study contribution patterns in academic vs general-purpose open source projects ### ❓ Questions - How can funding agencies better support research software infrastructure sustainability? - What interface design patterns work best for translating research tools to clinical applications? - How might research software projects better engage volunteer developer communities? ### 🔮 Future Research - Development of funding models specifically for research software infrastructure - Cross-domain conference impact on innovation outcomes - Clinical translation patterns for computational biology tools ### 💡 Implications - Practice: Cross-domain conference attendance should be standard for research software developers - Policy: Funding agencies need specific mechanisms for infrastructure vs discovery research - Education: Computer science curricula could include research software applications - Community: Regional conferences provide unique value for knowledge transfer between domains ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade #### 👀 Visual learners - Present: Protein folding visualizations mentioned, data plotting interfaces - Improve: System architecture diagrams, funding landscape infographics #### 🔊 Aural learners - Present: Podcast format with detailed explanations - Improve: Conference talk recordings, developer testimonials #### 🗣️ Verbal learners - Present: Clear definitions of MAVE, genomics terminology, funding concepts - Improve: Glossary of computational biology terms, analogy-driven explanations #### 🧩 Kinesthetic learners - Present: Django tutorial reference, hands-on development approach - Improve: Step-by-step MaveDB deployment guide, interactive variant lookup demo #### 🧠 Logical learners - Present: Clear problem-solution progression, funding challenge analysis - Improve: Decision trees for research software sustainability, metrics dashboards #### 🧑‍🤝‍🧑 Social learners - Present: Collaborative development model, community building focus - Improve: Community contribution guides, mentorship program details #### 🧘‍♀️ Solitary learners - Present: Individual reflection on cross-domain learning benefits - Improve: Self-assessment tools for research software readiness #### 🌿 Naturalistic learners - Present: Biological systems analogies (protein machines, cellular processes) - Improve: Ecosystem thinking about research software infrastructure ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure, descriptive episode title - Improve: Transcript formatting with speaker labels, time-stamped sections #### Keyboard navigation - Present: Standard podcast player controls - Improve: Keyboard shortcuts for timestamp navigation #### Contrast and readability - Present: Standard podcast platform compliance - Improve: High-contrast transcript options #### Media and captions - Present: Audio content with transcript - Improve: Chapter markers, searchable transcript segments ### 🌍 DEIB Assessment #### Cultural breadth - Present: Australia-US collaboration model, international standards work - Improve: Examples from diverse research contexts globally #### Inclusive language - Present: Accessible explanations of technical concepts - Improve: Avoid academic jargon without definition #### Neurodiversity support - Present: Structured conversation with clear topic transitions - Improve: Visual cues for topic changes, summary sections #### Representation - Present: Academic-industry collaboration perspectives - Improve: Diverse researcher voices, varied institutional contexts ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear infrastructure funding problem to specific solutions (20:00-25:00) - Inductive strength: PyconAU experience generalizing to conference value (42:00-45:00) - Causal reasoning: Open source tools enabling research capabilities (47:30) #### Fallacies - None identified: Conversation maintains logical consistency throughout #### Tone of voice - Overall tone: Conversational, educational, respectfully curious about cross-domain challenges ### 🛡️ DARVO Analysis - Deny: No denials present - open discussion of challenges - Attack: No attacks on accusers - collaborative problem-solving approach - Reverse victim/offender: No victim/offender dynamics present - Emotional cues: Genuine enthusiasm for cross-domain collaboration, realistic about challenges ## 🎯 Conclusion MaveDB exemplifies how research software can bridge multiple domains while facing classic infrastructure sustainability challenges. The project's success stems from building essential infrastructure, fostering cross-domain collaboration, and creative funding approaches. For the broader open source community, it demonstrates the value of regional conferences for knowledge transfer and the untapped potential for contributing to research applications that directly impact human health. Key takeaway: Research software sustainability requires treating infrastructure as essential utility rather than seeking direct monetization, while building strong community connections across domain boundaries. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 ### [H1] Cross-domain conferences accelerate research innovation > [!quote] [Episode 287: Alan Rubin on MaveDB](https://youtu.be/tEFJ9yqsCBw) by Sustain Podcast via YouTube. Published 2024. Accessed 2024-12-19. > "I went to PyconAU... saw a talk from Russell Keith McGee about Django... I told Doug that I could solve this problem" Regional conferences enable knowledge transfer between software engineering and scientific domains. [view highlight](https://youtu.be/tEFJ9yqsCBw?t=680) #xref #articleseed ID: --- ## Full Transcript [00:10] Hello and welcome to Sustain, the [00:12] podcast where we talk about sustaining [00:14] the open source ecosystem for the long [00:17] haul. Very excited to talk to my guest [00:20] today, Alan Rubin. I did skip the [00:22] traditional Gawk quote because I'm [00:24] wondering why I even quote Gawk at the [00:26] beginning of these podcasts. It's just [00:27] not necessary. Alan Rubin is a [00:29] computational biologist in Melbourne. He [00:32] has an open-source project called Mave [00:35] DB. Mavid DB. We'll figure out the [00:38] pronunciation later, but he did just nod [00:40] at Mave, so that's probably right. And [00:43] we're just going to talk today about [00:44] research software, open source, open [00:47] science, and how do we keep that ship [00:50] running and going on the endless tides [00:53] of the world. I am joined by no one [00:56] today. I'm the only podcast host. My [00:57] name is Richard Littower. I hope that's [00:59] okay with everyone. But of course, Allan [01:02] is on the line. Allan, how you doing [01:03] today? [01:04] >> I'm doing great, Richard. It's a [01:05] pleasure to be here. [01:06] >> It's a pleasure to have you here. So [01:08] Allan, as you may have already guessed, [01:09] O listeners, is originally from the US, [01:11] but now lives in Melbourne. And [01:13] apparently he's a group leader, which [01:15] kind of means assistant professor in [01:17] Australia, but he doesn't teach. And so [01:19] you just sort of run your own research [01:21] lab. Is that a correct understanding of [01:22] what you do? [01:23] >> That's right. I run an independent [01:26] research group in the collaborative [01:28] center for genomeic cancer medicine [01:30] which is a center that was set up [01:32] between the University of Melbourne and [01:34] the Peter McCllum Cancer Center which is [01:36] a very large cancer research hospital [01:39] here in Melbourne. And so we were really [01:41] set up to kind of drive things forward [01:44] in terms of precision oncology and using [01:47] genetics and genomics to help treat [01:49] cancer. And the center does a lot of the [01:52] genetic testing for cancer patients in [01:54] the state. And then we also have [01:56] obviously a significant research [01:58] component where we look at ways to [02:01] improve patient care as well as looking [02:04] at more fundamental research questions. [02:06] That is super cool. It's also way over [02:08] my head. I'm not an oncologist. I'm not [02:10] even sure what ank means in Greek. So [02:13] instead of focusing too much on that, I [02:15] want to point out to our listeners that [02:16] I met Allan when he and I were both [02:19] co-organizers of the academic track at [02:21] PYON AU. If you're interested in knowing [02:23] what Pyon AU is, that was the last [02:25] podcast with Jack, which was really fun, [02:27] where we talked about the regional [02:29] conferences that happen in this side of [02:30] the world and why regional conferences [02:31] are important. Both Allan and I are [02:33] interested in research software. So [02:35] Allan kind of has two hats. He's like a [02:37] cancer researcher on the one hand, but [02:39] an open- source wizard on the other. And [02:41] I think your main open source project [02:43] that you know about and work on would be [02:45] Mave DB. Can you tell me about Mave DB? [02:48] What it is and how it started? This is a [02:50] project to create a research database [02:53] for medical researchers, biological [02:56] researchers, biochemists, people that [02:59] are interested in molecular biology and [03:01] protein science and things like that. [03:03] And then we've recently expanded to try [03:07] to make the data much more accessible to [03:10] clinicians, so doctors and people that [03:13] are involved in clinical genetic testing [03:16] and interpreting the results of those [03:17] tests, which has been a big jump, at [03:19] least to me, and it certainly was a big [03:21] jump in terms of of what was needed. So [03:23] that's been kind of an interesting [03:25] journey. So if we try to remember back [03:27] to say high school biology, you may [03:29] remember that all of the activities that [03:32] are being carried out and all the cells [03:34] in your body are mediated by proteins. [03:36] So proteins are the little machines that [03:38] do everything inside your cells. And the [03:41] proteins function is determined by its [03:44] sequence. So it's a sequence of amino [03:45] acid that folds up into some shape that [03:47] then does a job. And that sequence of [03:50] amino acids that determines the shape is [03:52] in turn determined by the DNA sequence [03:54] that's in your genome. And this is why [03:57] if there is genetic change whether that [04:00] just happens like randomly and is passed [04:03] on from parent to child or happens in a [04:06] cancer cell and it changes the DNA which [04:09] leads to a change in the amino acid and [04:11] the protein which then changes how the [04:13] protein folds up. that protein then may [04:16] have a different function and most of [04:18] the time it just breaks. Like if you [04:19] think about if you had a complex little [04:20] machine like gears and clockwork and you [04:23] swapped out a part, the chances are it's [04:26] just going to catastrophically fail and [04:28] then you don't have that little machine [04:30] anymore which is probably bad if that [04:32] machine is responsible for some [04:35] important step of a more complex [04:36] process. But you might [04:38] >> like dependencies breaking in a long [04:40] chain of software dependencies, right? [04:42] is like you need the little time [04:43] dependencies back [04:44] >> if a bug gets introduced in the [04:46] dependency and of course we don't have [04:47] stack tracing tools in biology so it's [04:50] pretty difficult to figure out what's [04:51] going on so we essentially have to try [04:54] to figure it out by observing the output [04:57] of the entire system if you do make a [05:00] change the little machine might still [05:01] work but in a slightly different way [05:03] that protein may now bind to different [05:06] proteins or it might cause the cells to [05:09] grow much faster or make it so that it [05:12] now has no ability for the cell to stop [05:15] it and that's how you know sometimes [05:17] that can lead to cancer because you now [05:19] have a lack of regulation in the system. [05:21] So if what we need to do to understand [05:24] what these changes do because we can't [05:26] really directly observe what's going on [05:28] inside the cell. One of the ways that we [05:30] can figure out what these individual [05:32] changes do is we can make them in the [05:34] lab and test them. But now for each [05:36] amino acid you've got 20 different [05:38] options at each position. So now you've [05:41] got a 20N combinatorial problem which [05:45] means you have to be able to measure [05:47] many many thousands of these changes at [05:49] once if you want to measure all of them. [05:52] Now, fortunately, researchers about 15 [05:54] years ago developed a way to do this [05:57] where you can make essentially [05:59] synthesize DNA molecules, engineer them [06:02] into cells, make every different variant [06:05] of a protein, test it in a single pool [06:08] where they can kind of like compete [06:10] against each other or all be measured [06:11] like in parallel. And then you can use [06:13] high throughput DNA sequencing to [06:15] essentially count how many times you see [06:17] each variant when you started the [06:19] experiment and how many times you see [06:21] the variant at the end of the [06:23] experiment. And if you're measuring [06:24] something like the cell's ability to [06:26] survive with that variant in it, that [06:29] will give you a quantitative measurement [06:32] of [06:32] >> what I'm hearing is lots and lots of [06:34] data. So there's lots and lots of data [06:37] and basically what we're doing is to put [06:39] it in in software terms is we're [06:41] essentially doing fuzz testing in the [06:43] genome. [06:44] >> Cool. Okay. Awesome. [06:45] >> And then we're measuring like all of [06:47] those different little tweaks that you [06:48] can make and seeing whether it affects [06:50] the viability of the system. [06:52] >> So Mave DB is a database which allows [06:55] you to store all the different types of [06:57] variants and then to analyze them easily [06:59] to figure things out. [07:00] >> Yeah, that's right. So when we started [07:02] the project about 10 years ago, the [07:06] field had been around for 5 or so years. [07:08] There were starting to be quite a few [07:11] papers in this area that were describing [07:13] this very new kind of data where you've [07:15] done one of these assays. You've [07:18] measured thousands of unique variants in [07:21] a given protein. They all go together [07:24] and there wasn't really anywhere to [07:26] store the data. Obviously people would [07:30] share their data but it would go in a [07:33] supplemental table attached to a [07:35] scientific paper. So I don't know if for [07:36] people who are not familiar with the [07:38] academic literature when you publish [07:40] your paper in a journal there's like the [07:42] paper that describes the research and [07:44] then there often depending on the field [07:46] there are many many ancillary files [07:49] called the supplement which may include [07:51] extra detail for the methods or raw data [07:54] files or important technical information [07:58] that you would need to reproduce the [07:59] experiment. If you were lucky, there [08:01] would be a table of all of the scores [08:04] that were calculated for all of the [08:05] variants in the assays. And the issue [08:08] was that these were very difficult to [08:10] discover. They were all in very [08:13] different formats and it was kind of [08:16] getting out of control. And other fields [08:17] had databases to store all of the data [08:21] of a specific type and our field didn't [08:23] have that. And we thought that it might [08:25] be a good idea if we did. [08:26] >> When you say we, who are the people who [08:29] helped you out? Is this you and a [08:30] postgrad student? [08:32] >> Yeah. So I was collaborating very [08:33] closely and still do with professor [08:36] Douglas Fowler at the University of [08:37] Washington who actually started the [08:39] field and we've been friends for a long [08:41] time and we were talking about this [08:42] challenge and he had a meeting with Dan [08:46] Bolan at University of Massachusetts and [08:48] they were sort of talking about this and [08:50] they're like okay so you know we started [08:51] this thing and where is it going and how [08:53] are we doing and like oh yeah data [08:55] discoverability is like a real problem [08:57] and like we don't even know how many [08:59] assets there are and we can't get the [09:01] data and like all of this stuff. And me [09:04] being the guy that knew software who was [09:08] around thought that we could do this. [09:10] And actually, it's a nice kind of tiein [09:12] to your previous episode just the reason [09:15] why I took on this project was actually [09:17] because of Pyon AU because I went to one [09:20] of these regional conferences and I [09:24] remember I saw a talk from Russell Keith [09:26] McGee about Django and I never heard of [09:29] Django before but I was like wow sounds [09:32] like it's actually like not as hard as I [09:36] thought to develop like a datadriven web [09:38] application and So with great hubris, I [09:41] told Doug that I could solve this [09:43] problem and that I knew how to do it [09:45] using these open source tools. So there [09:48] was a guy Daniel Espazito who had just [09:51] finished his masters, didn't want to do [09:52] a PhD. He wanted to get like some job [09:54] experience. And so I pitched this idea [09:57] to him and I was like, "Do you know [09:58] about Django?" He's like, "Yeah, I've [09:59] heard of Django." I was like, "All [10:00] right, well, let's do it. So let's give [10:02] this a go." And so we quite literally [10:06] followed the Django girls tutorial, but [10:09] instead of building a blog, we built a [10:13] database for research scientists and it [10:16] worked very well. Like it was a very [10:18] capable system and Django lived up to [10:21] the promise of being this sort of like [10:23] batteries included kind of thing. And it [10:27] worked out quite well. we were able to [10:29] do all the things that we wanted and [10:31] because everything was built in Python, [10:33] we could bring in a lot of our expert [10:36] knowledge and we could, you know, bring [10:38] in tools like pandas to handle like data [10:42] frames and we could write custom [10:44] validators that were needed for like the [10:46] specific type of data that we had and we [10:49] could put all of that in there. So it [10:50] was relatively seamless like it in terms [10:54] of how we were able to approach the [10:56] problem and and how much of our existing [10:58] expertise we were able to bring. I did [11:00] like a deeper dive into like the history [11:02] of the project at Pyon AU a couple of [11:06] years ago and kind of went through the [11:09] ups and downs and the stops and starts [11:12] and struggles for funding and things [11:14] like that that go for any sort of [11:16] scientific open source project or really [11:18] probably any open source project. [11:20] >> I'm really glad that you mentioned [11:22] Russell Keith McGee. He's amazing. He's [11:23] been on this podcast. I dropped a link [11:25] in the show notes. Not to recap the Pyon [11:28] AU which is also linked in the show [11:29] notes Pyon AU talk you gave, but this is [11:32] great. You built a database, you built a [11:36] tool using open source tooling. Python [11:38] is such a rich ecosystem that you're [11:39] able to do a lot of stuff with it very [11:41] easily. And now we have this Mave DB [11:44] project. But this seems to me like this [11:46] is very clearly you are fighting cancer. [11:49] You're out there on the front lines [11:50] punching people and taking names and [11:52] being like cancer's got to go. That's [11:54] awesome. How is there not millions and [11:56] trillions of dollars for this project? [11:59] What are the struggles for funding for [12:00] Maved DB and why do they exist? It seems [12:02] to me a really clear use case when [12:05] you're doing something like building a [12:07] database within life sciences. There's a [12:11] lot of funding opportunities especially [12:12] compared to other academics like I don't [12:14] want to downplay how good we have it in [12:17] health and medical research but [12:19] essentially all of the funding [12:20] opportunities are oriented around [12:22] discoverybased or like hypothesis-based [12:26] projects and this is not one of those. [12:28] This is infrastructure and [12:30] infrastructure is just like [12:32] infrastructure in our public lives. [12:35] infrastructure is something that [12:38] everybody wants to have it exist and [12:40] nobody really wants to pay for it. And [12:43] so we face a lot of those same sort of [12:46] struggles in this area. There are [12:48] relatively few funding opportunities [12:51] directly for infrastructure and for [12:54] software infrastructure. Chan Zuckerberg [12:57] used to run one. I think they don't run [12:58] it anymore which is a shame. There are [13:00] occasional opportunities from larger [13:03] funders like the National Institutes of [13:06] Health, the National Human Genome [13:07] Research Institute historically has been [13:09] pretty good about funding these kinds of [13:12] projects directly. But those specific [13:15] funding calls are few and far between [13:18] and highly competitive because as you [13:20] can imagine, every database is trying to [13:22] get in on that. So I think that there is [13:25] a problem in that there just isn't [13:26] enough resources directed towards this [13:29] specific problem and it's generally a [13:31] bad fit for standard research grants. [13:33] >> So this is the classic infrastructure is [13:36] difficult to pay for problem that we see [13:38] across open source. We see it all the [13:41] time. Things like heartbleleed. It's [13:42] easy to get targeted money to fix it [13:44] from a company that's like crap. We need [13:45] that once it's a global headline. But [13:48] it's very difficult before then. you [13:50] can't do proactive work on your teeth [13:53] under most insurance schemes in the [13:54] world. Maybe that's a weird metaphor, [13:56] but like it's really really hard to do [13:57] this sort of stuff. Let's take that as a [13:59] given because I think we've covered that [14:00] in previous episodes of this podcast. I [14:03] think an interesting follow on might be [14:05] the data you're providing is really [14:08] useful for reproducing your research. [14:10] You mentioned earlier supplemental [14:12] tables. You mentioned papers that have [14:14] this sort of stuff. But when you're [14:16] providing a database that's incredibly [14:17] useful for ongoing work, what does it [14:20] look like in the field for funding that [14:23] sort of thing in the same way that we [14:25] now have universities fund APC charges, [14:27] publishing charges for papers in order [14:29] to have it be open, what about funding [14:32] for extra data and maybe funding for [14:34] infrastructural tools? Has that ever [14:36] been a good avenue for trying to find [14:38] marketing that Mave DB exist or funding [14:41] it directly? So, I definitely think that [14:43] there are people that are excited about [14:46] supporting the service. I mean, it's all [14:48] free. It's all open. We made a decision [14:51] very early on to make all the data [14:53] freely available, which allowed us to [14:56] grow, but it now means that we'd have [14:58] fewer options in terms of trying to get [15:02] financial support based on what it is [15:04] that we've built. We have had some [15:06] success adding support for the database [15:09] as part of larger research grants where [15:13] the dissemination of a large amount of [15:14] data is a critical aim of that grant. [15:18] And so that's been our strategy so far [15:21] and I think that's a reasonable way [15:22] forward. [15:23] >> A few months ago, maybe a year ago, I [15:26] had Eterna on this podcast which is a [15:28] large protein folding scheme where you [15:31] can have users go on and try to figure [15:32] out how proteins fold. This is kind of [15:35] early AI. I don't know whether it's all [15:36] AI now, but the idea is you can have [15:38] people figure out different ways of [15:39] protein folding, which seems relevant to [15:41] you in that it's in the space of medical [15:44] stuff and it's also in the space of [15:46] large open science, open-source project. [15:48] I'm curious if there's competition in [15:51] this field for that sort of funding. You [15:53] mentioned earlier a tiny thing where you [15:54] said the other databases are trying to [15:55] get in. Does it make it difficult to [15:57] find funding because you have to deal [15:58] with all these things? I think the [16:00] competition is difficult. If there's a [16:02] specific funding call that says, "Hey, [16:04] we want to fund data infrastructure." [16:06] Then everybody that does data [16:08] infrastructure is going to go and try to [16:10] get that funding. In which case, there's [16:12] competition at that point. I don't see [16:15] us as competing with really other [16:18] existing databases. Like we're [16:20] definitely not in competition for users [16:23] or in competition to be the place where [16:26] the data lives. We've been very [16:28] collaborative and open and have solved [16:31] some challenging problems. There's not a [16:33] long line of people that can do or [16:35] really want to take on the challenge [16:37] that we've taken on in terms of trying [16:38] to represent these data sets cuz it is [16:41] really challenging cuz I gave you a [16:43] simple example before, but there are [16:46] literally hundreds of different ways to [16:48] do these experiments. And so having the [16:50] right kind of structured metadata and [16:52] other stuff like that has been a heavy [16:54] lift. So we are in a really good [16:57] position in terms of people would much [16:59] rather link to us than try to replicate [17:02] what it is that we've done. And I think [17:05] that in terms of sustainability that has [17:08] been a way that we're thinking about it [17:09] because if we become a relatively [17:14] essential link in the sort of data [17:16] supply chain, if you will, then I think [17:19] it will be easier for us to rally [17:22] support if things get really dire. That [17:24] sounds similar to me to standardization. [17:27] So to to building a standard and having [17:29] people use the same standard, but that [17:30] also sounds to me like a lot of [17:32] community work. [17:33] >> Yeah, I would say that my role has [17:36] transitioned from being more technical [17:39] in the early days to being almost [17:42] entirely community focused and focused [17:45] on building an international network, [17:47] building a national network here in [17:49] Australia, engaging with everybody that [17:51] wants to use the data. And I think that [17:53] it's been quite interesting for me. So [17:55] when I first got into research and I did [17:58] my PhD, I had no interest whatsoever in [18:01] health applications or doing [18:03] translational research. I wanted to do [18:04] basic science only. If that was going to [18:07] have a positive impact on human health, [18:09] that's great. But somebody else was [18:11] going to figure that out in 25 years. [18:14] And it didn't have anything to do with [18:15] me. And I wanted to focus on really [18:17] fundamental problems. But then we [18:19] started this project and we got all this [18:21] data together and the clinical genetics [18:25] community came to us with like an [18:27] overwhelming amount of enthusiasm [18:30] because they were saying well we see [18:32] these genetic variants in patients and [18:34] we don't know what they do so we don't [18:37] know whether we can use them to help [18:40] guide diagnosis or treatment or anything [18:43] else. you've actually measured in an [18:45] experimental system what that variant [18:48] does and that is incredibly valuable [18:51] evidence for us to use when we're trying [18:54] to determine what that genetic variant [18:57] does and they're like and you have like [18:59] millions of measurements like you need [19:01] to give that to us immediately but of [19:04] course the way that they wanted to [19:05] experience the data was really different [19:07] to the way researchers wanted to we had [19:09] built a system where people would like [19:11] upload and download entire data sets [19:13] from a whole exper experiment, but they [19:14] really wanted to know, tell me [19:16] everything you know about this specific [19:19] DNA change. That's how they wanted to [19:21] experience the data. So, that kind of [19:24] started me on this process where over [19:26] the last several years, I've essentially [19:28] been learning all about clinical [19:31] genetics from field weers on conference [19:35] calls and trying my best to not offend [19:38] anyone with my ignorance. and and now [19:41] background in informatics. Was your [19:43] background not in [19:44] >> I mean my background is in computational [19:45] biology and genetics but much more sort [19:48] of like classical model organism [19:49] genetics and [19:51] >> got it [19:51] >> kind of computational biology [19:53] mathematical modeling kind of stuff. So [19:54] nothing clinical once you start talking [19:56] about clinical applications like all the [19:58] vocabulary changes the mission changes [20:02] cuz it's not about this kind of abstract [20:04] idea of creating new knowledge for [20:06] humanity which is laudable. It's very [20:09] much about how can we help individual [20:12] people and what is very motivating to me [20:16] talking to my clinical colleagues is [20:19] that when they're thinking about these [20:22] problems like they have this very real [20:24] experience of a person who needs their [20:29] help who is sitting in front of them who [20:31] is like the reason why they are doing [20:33] this. because when they're trying to [20:35] diagnose a genetic change in a patient [20:38] like that's a real person for them and [20:40] it because of my background that was [20:41] very abstract and so that has been quite [20:44] interesting to kind of think about those [20:45] problems in a different way but I will [20:47] say that it took the project in a whole [20:50] different direction because they needed [20:52] a completely different interface we [20:54] needed to try to come up with some [20:57] answers to some challenging questions [20:58] about like data quality which again very [21:02] challenging because the field was so [21:04] varied in terms of what kinds of things [21:06] they're doing, what kind of experiments [21:08] they're performing, you can't really [21:10] directly compare them. So yeah, so that [21:12] took us in a completely different [21:13] direction and we've now launched a new [21:17] interface for the MADB database which [21:19] we've called Mave MD which is Maves for [21:21] Medicine which is an interface that is [21:24] entirely designed for clinical [21:27] geneticists who want to learn more about [21:29] a specific variant and learn about [21:32] whether they can use it to help resolve [21:34] their patient case. And so that was a [21:36] huge pivot for us. Again, the existing [21:40] library and sort of richness of open [21:42] source tools enabled us to do that. [21:46] >> What I'm hearing is that you started as [21:48] a computational biologist, you run your [21:50] own research lab, you run your entire [21:52] own source project with a couple [21:54] collaborators and now you're learning a [21:57] whole new field and catering to the new [21:58] field. And so what I'm not hearing is [22:00] when you sleep. And so I'm curious about [22:02] the institutional changes you've made to [22:05] your open source project to enable you [22:07] to sleep over the past few years because [22:10] it sounds like a ton of work and I'm [22:11] just wondering how have you made that [22:12] sustainable? [22:13] >> Yeah, it is a ton of work and I think [22:15] that what's made it sustainable is that [22:17] we have really great support from my [22:19] close collaborators at the University of [22:21] Washington where much of the software [22:23] team that works on the project is based. [22:26] Some of us are in Melbourne, some of us [22:28] are in Seattle. It's been really great [22:30] to have this very experienced team come [22:33] together and work on this project and [22:35] I'm super grateful to have them. It [22:37] means that I don't need to be [22:38] responsible for the day-to-day technical [22:41] decisions. I think if I were in a tech [22:45] company context, I would basically be a [22:47] product owner in that I am sort of [22:50] directing the highle goals of the [22:53] project and I know meeting with [22:56] stakeholders feels weird saying that but [22:59] you know reaching out to you know other [23:01] researchers and essentially spending a [23:03] lot of time on Zoom at inconvenient [23:05] times for me cuz Australian time zones [23:08] are not friendly to having a meeting [23:09] with people in North America. and Europe [23:13] and me all at the same time. [23:15] >> Let's not even go there. It's the worst. [23:17] I had a really good podcast a couple [23:19] years ago with Tracy Hines. This was [23:21] episode 135 where we talked about how [23:23] open source needs more PMs. And I think [23:26] we meant project manager, but there's [23:27] also product manager and product owner. [23:31] I guess POS would be another variant of [23:33] that, which is someone who oversees the [23:35] whole thing. Do you have any [23:37] institutionalized distinctions between [23:38] roles in your project or is it just core [23:41] maintainers and then you have like a [23:43] bi-weekly call or something or is there [23:45] some setup you have where people can set [23:48] into these specific roles that may be [23:50] useful for the community? So I think [23:51] like a lot of academic projects [23:53] everyone's wearing a lot of hats all the [23:55] time and we don't have the luxury of the [23:58] headcount to have a different person who [24:01] could be a project manager. I'd say I'm [24:03] not a project manager because I'm not [24:05] really responsible for setting timelines [24:08] and deliverables and things like that. [24:09] It's more setting the highle agenda and [24:13] saying like this is something that we [24:14] ought to prioritize or I talked to these [24:17] folks and they really want like a [24:19] feature that does this or using my own [24:23] knowledge of the field and the history [24:25] of the field and the project and [24:26] everything like that to say well this is [24:28] something we should prioritize or this [24:29] is something else that we should dep [24:30] prioritize. So that's a lot of my role [24:33] now and you know going to conferences [24:35] engaging with people and things like [24:37] that. We have a team of four people. So [24:40] we have a senior developer whose primary [24:42] job it is to work on Mave DB and then we [24:45] have other people who contribute in [24:47] their roles as they can. We don't really [24:50] get community contributions like it's [24:52] just not that kind of project. This is [24:55] something that I had always thought [24:59] would be easier to sell to the broader [25:01] open- source community, but I think it's [25:03] a combination of open source hobbyists [25:05] are intimidated about contributing to a [25:08] project like this that has this kind of [25:10] like realworld application. I think we [25:14] as people that run the project don't [25:17] really know how to engage with hobbyists [25:20] and we aren't really doing the usual [25:23] like big open source project thing like [25:25] tagging issues as good for new [25:27] contributors and leaving them there. [25:29] We're just very focused on the mission [25:32] and like getting stuff done. I think the [25:34] other thing is that when people are [25:35] choosing a hobby project, they want [25:37] something that is like maybe more fun or [25:40] that's more related to like a hobby that [25:42] they have and while they may be strongly [25:45] aligned with like the mission that we [25:47] have like you know we're trying to cure [25:49] cancer is like pretty compelling story [25:52] that's ultimately like not where they [25:55] want to spend their hours when they [25:56] could you know like build a robot or [25:58] something like that. Honestly, I don't [25:59] blame them. The dynamics of the project [26:00] are it's much more like a kind of [26:02] commercial software project and that [26:04] like we have a team of people that are [26:06] paid salaries to work on it and we do [26:08] that. I have never been successful in [26:10] engaging the broader open source [26:12] community despite having tried many [26:14] times to contribute to this or any other [26:17] scientific project. Although we sure [26:19] could use the help. [26:20] >> Parts of my brain are screaming at the [26:22] idea of the broader open source [26:23] community existing as a as a unified or [26:26] useful group to talk about. Also the [26:28] idea of hobbyists being the large part [26:30] of that. I think the majority of open [26:31] source people are actually students or [26:35] people who are paid by their company to [26:36] do the work. That's a gut feeling. [26:38] That's not that useful. I hear you on [26:40] how difficult it is to get hobbyists to [26:42] contribute to research software though. [26:43] That is tough. [26:44] >> Yeah, I think that's probably correct. [26:46] But I mean I'm thinking you know when I [26:47] have gone into spaces like Pyon AU which [26:50] is full of people who code all day at [26:53] work and then go home and code more on [26:55] something else which is not how I'm [26:57] wired but I love that those people are [27:00] out there just doing it. trying to get [27:03] them to adopt a project from like a [27:05] researcher that has all the domain [27:07] knowledge and none of the software [27:10] engineering expertise or insufficient [27:11] software engineering expertise and like [27:15] help them out. If we could somehow [27:17] figure out how to connect those people, [27:19] I think it would be very helpful and [27:22] probably really fulfilling. You need to [27:24] talk a bit to me about that actually [27:26] because I often find when I go to [27:28] conferences that are slightly outside my [27:30] field, they're interesting in the sense [27:32] of like, oh, that's kind of cool, but I [27:33] very rarely feel like they're that [27:35] beneficial to me. Tell me why that's not [27:37] the case for you because a lot of this [27:39] started you said at Pyon AU, but those [27:42] are different sets of people like what [27:44] is it that you get from going to [27:45] industry conferences when you're an [27:46] academic? [27:47] >> Yeah, that's a really good question and [27:49] it's something that I feel very strongly [27:51] about. So if I as a computational [27:53] biologist go to a biology conference, [27:56] I'm going to see a bunch of new science. [27:59] I'm going to learn a lot of new things, [28:00] but it's all going to be applications. [28:02] I'm going to learn new applications, new [28:05] areas that I could get into, new things [28:08] that I could study. I'm not going to [28:10] learn anything about how to do what I do [28:14] on a day-to-day basis, which is mostly [28:16] write software or do data science. I'm [28:19] not going to learn how to do those [28:20] things better. It's not a way for me to [28:22] work on my craft in a way that it is for [28:25] people that are at the bench doing like [28:28] sort of wet lab experiments where they [28:30] would learn about new techniques at one [28:31] of those biology conferences. I need to [28:34] go to where the software engineering and [28:36] data science people are talking about [28:38] how to improve their craft and that is [28:41] places like Pyon AU. I went to Linux [28:45] Conf in Australia back when that was [28:47] still running and that was where I [28:49] really learned a lot about things that [28:51] are kind of basic concepts probably for [28:54] a lot of your listeners like things like [28:56] how to write a unit test suite that [28:57] isn't a mess, right? I remember there [29:00] was like a a long tutorial talk about [29:03] like how to actually write a good unit [29:06] test and it talked about using standard [29:08] library unit test versus using piest and [29:11] it talked about test runners. I did [29:14] >> most of like an undergraduate computer [29:16] science degree. You don't learn about [29:17] that there either. And so one of the [29:19] things that I've really tried to [29:21] convince my colleagues of is that going [29:24] to something like Pyon AU is incredibly [29:29] valuable in terms of making your science [29:31] better because you will learn more about [29:34] open source tools that can help you be [29:37] more efficient. And I think that [29:39] especially as we've seen the rise of [29:42] data science in industry, we were doing [29:44] data science the whole time, just nobody [29:46] called it data science cuz nobody was [29:47] really talking about it. We had large [29:49] data frames with like millions of rows [29:52] and thousands of columns and we had to [29:54] figure out how to like do stuff. We were [29:56] there the whole time plugging along in [29:58] pandas when Excel broke and now there's [30:02] all of this great support from the open [30:04] source community. And so going and you [30:06] know hearing about when should I use [30:08] pandas or polars or something that's [30:12] based on arrow or like any of the other [30:14] you know proliferation of options that [30:16] are there. I think it that's something [30:17] that more scientists need to get exposed [30:19] to and you know these open source [30:21] conferences and open source meetups and [30:23] things like that. That's the place to go [30:25] to do that. [30:26] >> On that note you don't have to go to [30:28] Pyon AU people. You can also go to [30:30] Python EU. Go to the Python you know go [30:32] to the conferences where you are. You're [30:34] totally okay. I have a different [30:36] question, Alan, which is what can the [30:40] normal open-source practitioners that [30:43] are at these conferences that listen to [30:45] this podcast that do open source on [30:46] GitHub? What could they learn from you [30:48] as a scientist that you don't think that [30:50] they know very well? Is there something [30:52] else that you can give back? No is okay. [30:54] I mean, it's okay if you have no idea. [30:56] I'm just I'm just curious if there's [30:58] some correlary where they should be [30:59] going to biology conferences. I think [31:01] people that work in open source should [31:03] really be excited about the kind of [31:05] stuff that their work is enabling even [31:08] if they don't really hear about it. I [31:09] think that for every large company [31:12] website that is running using their [31:14] tools, there are hundreds of other like [31:17] little academic projects that only exist [31:20] because all of this stuff is freely [31:22] available and well doumented there for [31:25] people to leverage. So like we started [31:27] on Django, then we switched over to fast [31:29] API. I mean it was a full rewrite but it [31:32] was a relatively straightforward process [31:34] because these tools are like very mature [31:36] and very well documented. You know we [31:39] have no resources in academia to go and [31:42] like purchase software license software [31:44] like in general everything that I use [31:47] that's licensed is licensed at the level [31:49] of my institution. So that's things like [31:51] Microsoft Office. [31:53] >> They're not going to buy anything for me [31:55] and I'm not going to be able to request [31:58] it. So I think that it is hugely [32:00] enabling and people should be excited [32:03] about that and consider you know [32:04] checking out public lectures and things [32:07] like that other sorts of ways to engage [32:09] with their local scientific community. I [32:12] think another way that people can get [32:14] involved that would be super valuable to [32:16] scientists is that especially in [32:18] Australia there's a large focus on what [32:21] we call healthcare consumers which are [32:24] people from the general public who are [32:28] engaged in or maybe their family has [32:30] been affected by or they're personally [32:32] affected by some particular condition [32:36] and their role is to talk to the [32:38] researchers that are working in that [32:40] area and give them advice and talk to [32:43] them about their stories and help them [32:46] write grants and refine their research [32:49] so that it's actually going to help the [32:51] people that it's supposed to help. So [32:52] it's not quite so abstract and I think [32:55] that it can be a struggle for people who [32:59] are doing software oriented work to find [33:03] a good match with a healthcare consumer [33:06] because most of them are not technical. [33:08] But I have a great consumer who I've [33:10] been working with who does have a [33:12] background in like enterprise IT and [33:16] she's fantastic and can give me a lot of [33:18] really good advice on the kind of stuff [33:20] that I'm working on. But I think that if [33:22] more folks from the open source [33:24] community were interested in getting [33:25] involved with these programs, I would [33:27] hope it would be very rewarding for them [33:28] and also super valuable for the [33:30] researchers that are doing computational [33:32] work. [33:32] >> I like that sentiment. This is a really [33:34] fun conversation. Alan, we are running [33:36] up on time. So, I just want to thank you [33:38] for having it with me and for talking [33:41] about Mave DB and how it's been going [33:43] and how it's useful. Listeners who are [33:45] curious to see more about the [33:46] sustainability should check out the uh [33:48] link to the talk at PONA AU. I think [33:50] it'll probably be similar to the stuff [33:51] we've talked about before in this [33:53] podcast, building community, doing the [33:55] long work, putting in the hard hours, [33:56] trying to find funding, etc., etc., etc. [33:59] But Alan, I appreciate you coming on and [34:00] talking a bit more about this sector of [34:02] open source that is super important, [34:04] super relevant and all of us end up at [34:07] some point in a hospital. So super, you [34:09] know, personal. This is this is great [34:11] stuff. Don't leave yet though. First [34:13] off, I want to ask where can people [34:15] learn more about Mave DB and your work? [34:19] >> So people can go to our website which is [34:21] mavendb.org and they can have a look. [34:24] I'm not sure how comprehensible it will [34:26] be, but we do have interesting looking [34:28] data visualizations. You can figure out [34:30] how to plot things on a protein [34:32] structure and kind of spin it around, [34:34] which is good times. Much of the work [34:38] that we've done is in the scientific [34:39] literature. There are some pieces that [34:41] are more accessible to like a general [34:44] scientific audience, and I can provide [34:46] some links to those papers. We're also [34:48] part of an international kind of [34:51] volunteer oriented group called the [34:53] Atlas of Variant Effects Alliance, which [34:55] is bringing together the people that are [34:57] generating all of these data sets, [34:59] wanting to use the data sets, funders, [35:02] doctors, researchers, people across the [35:06] spectrum to try to build this field. And [35:09] that's an organization that I've been [35:10] involved with for a long time. I'm on [35:12] their executive committee. I lead weed [35:13] one of the work streams. They're doing [35:14] really good work. And so that's an [35:16] another place that people can go to just [35:18] kind of learn more about what is this [35:20] field trying to do like what's the [35:22] ambition and what are we all about. [35:23] >> That's variant effect.org or maved db [35:26] that's mavved db.org. Thank you so much [35:30] Alan. Don't leave yet. At the end of [35:31] every episode we have spotlight. [35:32] Spotlight is the part of the show where [35:34] we just sort of change tac, go to left [35:36] field and talk about people, things, [35:38] persons, whatever projects that we think [35:40] need a bit of light shed on them. [35:44] Traditionally, the host goes first. The [35:46] second most recent book that I have [35:47] finished is called News of the Dead by [35:50] James Robertson. James Robertson is a [35:54] Scottish author who writes about [35:56] Scotland and what it means to him in [35:58] general. But News of the Dead is a [36:00] fantastic novel about a Glenn up way, [36:03] which sort of covers a thousand years [36:05] worth of history in this Glenn. And it's [36:07] one of these excellent books that makes [36:08] you stop, think, and really want to live [36:11] in a single place for the rest of your [36:13] life. Um, so highly suggest it, [36:15] especially if you like hear than [36:16] anything written in Scots. It's not [36:18] quite eye dialect. It's a totally [36:20] different language. Super super fun. Do [36:21] check it out. Alan, what is your [36:24] spotlight today? And keep in mind, it [36:25] doesn't have to be a book. I've just [36:27] done hundreds of podcasts and I'm out of [36:28] those open source software packages. [36:30] What's what's yours? Well, I talked a [36:32] little bit about how important the sort [36:34] of broader Python ecosystem and the [36:36] Python web development ecosystem has [36:38] been for this project, but I think I [36:40] want to spotlight something which we [36:43] didn't really get into, which is the [36:44] importance of open data standards and in [36:47] particular the work that's being done by [36:48] a group called the Global Alliance for [36:50] Genomics and Health or G4GH which is an [36:53] international volunteer body for [36:56] maintaining and evangelizing data [36:58] standards for genomics and It's really [37:01] important work, very unglamorous, but [37:04] the standards that they're developing [37:06] are completely essential for all of the [37:09] stuff that we would want for our this [37:11] sort of digital health future of which [37:13] much has been written about and which [37:15] we're trying to realize, but it's very [37:16] hard. And this is essentially what are [37:18] the data standards that are going to [37:20] allow APIs to securely and in a [37:24] privacy-minded way share health [37:26] information about individuals between [37:29] providers and hospitals and subsets of [37:33] it with researchers and allowing all of [37:36] these different systems to talk to each [37:38] other. You know, somebody's got to do [37:40] that work. Doing it across national [37:42] boundaries is incredibly challenging. [37:45] doing it in a in an environment that is [37:48] as complicated as biology and human [37:51] health is incredibly challenging. And so [37:53] I just wanted to spotlight the good work [37:56] that they're doing. [37:57] >> I love that. Thank you so much. You can [38:00] find the link to that project and all [38:02] the other things that we've mentioned [38:03] including some things we haven't. For [38:04] instance, I forgot to talk about Karen [38:06] Sandler who basically also started from [38:08] a health open source perspective. Super [38:10] super cool. Online at the show notes you [38:12] can go to sustainoss.org org or [38:15] podcasts.org [38:16] to find the show notes for this episode [38:18] and other episodes like it. Listeners, [38:20] if there's other things that move you [38:23] and you would like to do more, there's [38:24] all sorts of things you can do to help [38:26] this podcast out. One of them is share [38:27] it with your friends, enemies, and [38:29] neighbors. That would be really, really [38:31] dope. Talk about the podcast, put it on [38:33] your social media. You can also like us [38:36] on Apple Podcast, Spotify, etc. Wherever [38:39] podcasts are bought and sold and vended. [38:42] Follow Sustain on Mastadon and LinkedIn. [38:45] You can go to our Discourse on [38:46] Discords.org. [38:48] You can email Richard at [38:50] sustainosess.org. If you're really, you [38:52] know, interested in sending your hate [38:54] mail my way or love mail, whatever you [38:56] want. Not too spicy. Please, if you have [38:58] any feedback, you can also email all the [39:00] hosts at podcasts.org. [39:02] And most importantly, you can donate to [39:05] keep this podcast going. We have some [39:07] funding for the next year, which is [39:08] really cool. It is a bit more academic [39:11] focused which is why I'm having another [39:12] academic podcast. I also just secretly [39:14] really like them. I think it's a part of [39:16] the open source ecosystem that's really [39:17] really important and it's no less valid [39:19] than people working in the kernel. But [39:21] you can go to open collective and go to [39:23] sustain oss and donate to this podcast [39:25] where it's running costs on average it [39:27] costs around $175 per episode. So it [39:30] would be super dope to have that money [39:32] reimbured. And besides that again just [39:35] you know spread the love. Go check out [39:36] Mave DB. Say thanks to Allan online. You [39:39] can find his socials on the show notes [39:41] as well. And otherwise, have a good day. [39:43] And likewise with you, Alan. Thanks for [39:45] coming on. Um, really appreciate it. [39:47] Thanks for having me. This was a lot of [39:48] fun.  📄 **Full text:** ![[Full Document Contents/Articles/Episode 287 Alan Rubin on MaveDB.md|Episode 287: Alan Rubin on MaveDB by YouTube]] ## Highlights - the world and why regional conferences are important. Both Allan and I are interested in research software. So Allan kind of has two hats. He's like a cancer researcher on the one hand, but an open- source wizard on the other. And I think your main open source project that you know about and work on would be Mave DB. Can you tell me about Mave DB? What it is and how it started? This is a project to create a research database for medical researchers, biological researchers, biochemists, people that are interested in molecular biology and ([View Highlight](https://read.readwise.io/read/01kpgxbs0cshb2szs85s0bcrr1)) - Tags: 👻 ai highlighted - great. You built a database, you built a tool using open source tooling. Python is such a rich ecosystem that you're able to do a lot of stuff with it very easily. And now we have this Mave DB project. But this seems to me like this is very clearly you are fighting cancer. You're out there on the front lines punching people and taking names and being like cancer's got to go. That's awesome. How is there not millions and trillions of dollars for this project? What are the struggles for funding for ([View Highlight](https://read.readwise.io/read/01kpgxbs1qmzysp65abfywg8ef)) - Tags: 👻 ai highlighted - can have users go on and try to figure out how proteins fold. This is kind of early AI. I don't know whether it's all AI now, but the idea is you can have people figure out different ways of protein folding, which seems relevant to you in that it's in the space of medical stuff and it's also in the space of large open science, open-source project. I'm curious if there's competition in this field for that sort of funding. You mentioned earlier a tiny thing where you said the other databases are trying to get in. Does it make it difficult to find funding because you have to deal with all these things? I think the ([View Highlight](https://read.readwise.io/read/01kpgxbs3mgq8m0wxzyceb8nd3)) - Tags: 👻 ai highlighted - whether they can use it to help resolve their patient case. And so that was a huge pivot for us. Again, the existing library and sort of richness of open source tools enabled us to do that. >> What I'm hearing is that you started as a computational biologist, you run your own research lab, you run your entire own source project with a couple collaborators and now you're learning a whole new field and catering to the new field. And so what I'm not hearing is ([View Highlight](https://read.readwise.io/read/01kpgxbs6c86teex2wpzgef6m0)) - Tags: 👻 ai highlighted - are it's much more like a kind of commercial software project and that like we have a team of people that are paid salaries to work on it and we do that. I have never been successful in engaging the broader open source community despite having tried many times to contribute to this or any other scientific project. Although we sure could use the help. >> Parts of my brain are screaming at the idea of the broader open source community existing as a as a unified or useful group to talk about. Also the idea of hobbyists being the large part ([View Highlight](https://read.readwise.io/read/01kpgxbsdbgg2ftc2srhhwshgk)) - Tags: 👻 ai highlighted - researchers that are doing computational work. >> I like that sentiment. This is a really fun conversation. Alan, we are running up on time. So, I just want to thank you for having it with me and for talking about Mave DB and how it's been going and how it's useful. Listeners who are curious to see more about the sustainability should check out the uh link to the talk at PONA AU. I think it'll probably be similar to the stuff we've talked about before in this podcast, building community, doing the long work, putting in the hard hours, trying to find funding, etc., etc., etc. But Alan, I appreciate you coming on and ([View Highlight](https://read.readwise.io/read/01kpgxbsp8ctwx0wbxschxj2cx)) - Tags: 👻 ai highlighted - it with researchers and allowing all of these different systems to talk to each other. You know, somebody's got to do that work. Doing it across national boundaries is incredibly challenging. doing it in a in an environment that is as complicated as biology and human health is incredibly challenging. And so I just wanted to spotlight the good work that they're doing. >> I love that. Thank you so much. You can ([View Highlight](https://read.readwise.io/read/01kpgxbt0dhx12p88jgy930rje)) - Tags: 👻 ai highlighted