Golgi Turns One: A Year of Innovation and Growth Today, Golgi celebrates its first anniversary! Reflecting on the past year, it’s clear that when every day presents a new challenge, a year can feel like a lifetime. It’s only been 12 months, but the journey has been filled with unforgettable experiences. Starting a business is an adventure, with exhilarating highs and challenging lows. Entrepreneurship is far removed from the comforts of a salaried job. Paychecks are uncertain, and each day is a test—perform or fail. The pressure can be intense, but the highs—methodological breakthroughs, landing our first customer—have been extraordinary. Golgi began as a simple idea—a daydream about what I could build. But once the wheels started turning, Golgi took on a life of its own. Colleagues from every stage of my career—from CU-Mathematics, UNC-Biostatistics, HMS-Cell Biology and Calico Life Sciences—reached out, eager to join the adventure. Thanks to a mix of serendipity, luck, and determination, we were able to bootstrap a business with no outside funding. I owe an immense debt of gratitude to our founding team. They didn’t have financial incentives—there was no money to drive them—yet they showed up and delivered, day in and day out. How many people would continue working if their salaries vanished, with only the love of the work and the potential for future rewards to keep them going? Not many, but that’s exactly what our team did. Every member believed in our mission and trusted each other to get the job done. Working with such a dedicated and passionate team has been a privilege I’ll cherish forever. Bootstrapping a business comes with its own set of challenges. Money is always tight, and every decision feels critical, adding layers of pressure. It’s not a path for everyone, but it worked for us. The financial constraints forced us to stay laser-focused on what mattered most—solving real problems for our early customers. Without the distractions of fundraising or job hunting, we dedicated ourselves to what we do best—building cutting-edge tools. The results have been nothing short of remarkable. Initially, we offered solutions to niche problems like eliminating rare outliers and improving performance in complex experimental designs. But thanks to R&D, we developed a transformative value proposition: we could double assay throughput with no loss in performance! This shift resonated with customers, quickly turning leads into paying clients. In our first 12 months, we built a powerful data analysis platform, initiated two patents, and secured an early customer base. There’s remains plenty of uncertainty the future of our young enterprise, but it’s hard to imagine a better start. To everyone who offered advice and encouragement throughout the year, thank you! Your words have made a difference and I’m looking forward to sharing many more milestones with you in the years to come. The journey has just begun! Jonathon O’Brien www.golgi.bio
关于我们
We are the computational Golgi, modifying your data and placing it exactly where it needs to be.
- 网站
-
https://www.golgi.bio/
Golgi的外部链接
- 所属行业
- 生物技术研究
- 规模
- 2-10 人
- 总部
- San Francisco
- 类型
- 私人持股
- 创立
- 2023
- 领域
- Proteomics、Bioinformatics、Software和Analytics
地点
-
主要
US,San Francisco
Golgi员工
动态
-
Golgi转发了
Eric Dai and I wrote an essay on the impact of AI/ML in drug discovery and why so much discourse on the topic is confused. Check it out! https://lnkd.in/eFcayiym
-
Good people. Go work with them.
Xaira has an ambitious plan to generate rich Proteomics and Functional Genomics datasets that propel drug discovery through application of AI/ML. Today we are opening 4 new positions to expand the impact of Proteomics...from elucidating disease biology and supporting drug discovery programs to establishing technologies on the analytical and computational edges of our field. Join us! https://lnkd.in/g5gY4MN8 https://lnkd.in/g5uy6sdq https://lnkd.in/ggzQu3jG https://lnkd.in/gv8t-WAn
-
Golgi转发了
Nice post from David. Some thoughts... Efforts to create foundation models in biology quickly end up touching on the field of preclinical decision science. Ideally, discovery efforts are devoted to experiments with the highest predictive validity, eg. those that score therapeutic targets/candidates in a way that ultimately correlates well with clinical utility. Many therapeutic areas use models with poor (or unmeasured) predictive validity, so the duty cycle to understand whether any new discovery platform is performant is incredibly long. Probably the fastest way for an AI newco to go from foundational model--> traditional value-generating point in biotech would be to point their efforts initially towards clinically validated surrogate endpoints. As an example, LDL levels are monotonically associated with cardiovascular disease risk, so it would be compelling and fast for a newco to rapidly go from foundational model-->new CVD target--> actually compelling preclinical POC data. David's post also touches on the business model/economic considerations that foundation models in both tech and biotech have been grappling with: a) is there any defensibility to the business? Are foundation models a real catalyst to build a business around (or something closer to a science project)? b) Will benefits accrue to incumbents or newcos? c) is this a cheap parlor trick or a real innovation that solves real business problems better/faster/cheaper than the best alternative options?
Computational Biology, AI/ML and Biotech. I specialize in bridging complex biological problems and computer science solutions. Ex-Altos Labs | Ex-eGenesis | Cornell BS| Caltech PhD | MIT scientist
More and more often, I am seeing computational biology decks that worry me. Why? Because they all look identical: they propose that biology is complicated (gasp!), and they posit that to solve it we need better AI tools that make sense of data. Inevitably, these decks call for multimodal, high throughput (single cell) data, and they make calls for the development of foundation models. They end by showing promising slides where they can explain a bit of the overall variance in the data, and claim that this progress will lead to much faster advancement in drug development. What's wrong with this? Why does this worry me? Let's be clear--the reason is not that AI/ML is bad or wrong. What worries me the most about these slide decks is that they are interchangeable among companies. Not a single slide deck I have seen contains ideas or approaches that are unique to one company over another. The ideas are clonal--therefore, if one deck is flawed, chances are they are all flawed. Moreover, if everyone has the right approach, then the winner will be determined by dataset size, which means if you are a small or new company, you're inherently at a disadvantage by choosing to take the exact same approach as big pharma. Computational biology and AI/ML are big fields, so why are we seeing total convergence of ideas and approaches? This leads me to my next concern. All the decks I'm seeing lament the complexity of biological data, and argue that we need to improve models to increase the rate of development in biology. However, not a single one of these decks tells us how development will be augmented by AI. One common claim is that these models will enable perturbation prediction--but this is a non-starter. Why? Because most drugs act to modulate physiology, but these models are trained on in vitro cell culture data. What does it mean to predict perturbation effects in fibroblasts when the drug acts on, say, Kupffer cells, for which we have practically zero data in vitro? How and why is this superior to an empirical drug screen? How and why is it cheaper (recall: these models require enormous amounts of GPUs to run). I have found that when we ask ML engineers how we will validate their models, the answer is typically vague and hand-wavey, and often involves long timescales. I argue that this is anathema to computational biology. Call me old fashioned, but if you give me a good experiment, some data from that experiment and a bit of PCA, I can generate testable hypotheses in just over 20 minutes. Why is the bar for these AI models that their effects on biological research will be validated over the next decade? It is important we insist on strong tests of deep learning models. A strong test is NOT a test of generalizability or on-task performance. Foundation models should generate immediately testable biological hypotheses that cannot be reached using other methods. Let's validate AI not on variance explained, but on signal extracted. #ai #ml #compbio
-
Golgi转发了
Wondered how to estimate absolute amounts of glycopeptides by mass spectrometry without using any standards or calibration curves. Check out how ! "Standard-free quantification of glycopeptides via coulometric mass spectrometry" that appeared as cover article in JASMS https://lnkd.in/gyPcBF6U
-
Golgi转发了
Couldn’t have articulated this better myself. And shameless plug for some of the ones I’ve found particularly innovative of late if you’re looking for cool solutions - Golgi for mass spec data (#datanerdalert re how much I love the technical details of this one for this data type, and from truly deep experts in the field), and Ontologic for the simple flexibility they bring to data of all sizes.
It's a common perception that there's a lack of good software for biotech. But the real problem seems to be that teams don't bother looking for it. Sure, there are a number of notable, widely-used pieces of software that are overpriced and look like they haven't been updated in the last decade (or two). But there are many more examples of modern, usable, reasonably priced options out there that aren't getting nearly the attention they deserve. I think we've done such a good job convincing ourselves that biotech software is bad that we don't bother looking for better options. Instead, we stick with the handful of "industry standard" options that feel safe, or we build internal systems from scratch. And when others ask for suggestions, we tell them there's no good options. Well, I think it's time for this to end. We need to start shifting the narrative: There's good options out there. We just don't have great ways to find it (yet). Oh, and we also have to come up with better ways to find it... I wrote about this in more detail in the latest post to my weekly newsletter, Scaling Biotech. https://buff.ly/4bmlNK8
Biology software isn't bad. You just haven't found the good stuff.
scalingbiotech.substack.com
-
Golgi's first ASMS conference ended all too quickly! I had a fantastic time reconnecting with old colleagues and learning about the latest emerging technologies. But more than anything else, I loved introducing Joseph Newhall to his first ASMS. Joe, who comes from a pure mathematics background, was blown away by the conference. He had no idea how much intellectual activity was happening just in the field of proteomics methodology. Experiencing it through his eyes made me feel ten years younger and gave me a renewed perspective on the mass spectrometry community. The ingenuity and sincere love of discovery that permeates every corner of ASMS is truly remarkable. It’s a community I am proud to be part of and a gathering I will always cherish attending. See you all next year! Jonathon #ASMS2024
-
Golgi is the hottest sub cellular component at ASMS and it’s not even close! Were the other organelles even trying? Kevin Dong Thao Nguyen #ASMS2024