Sphinx Bio

"Can you rerun this analysis in Python?" If you're a bench scientist, these words probably make your heart sink. If you're in computational biology, you've probably said them more times than you can count. And if the end of that sentence is R instead of Python, then everyone is unhappy... This disconnect between bench and computational teams isn't just frustrating - it's costing biotechs precious time in their race to develop new drugs. The truth is, most routine biochemical assay analysis follows standard patterns. It shouldn't require either manual spreadsheet work or custom coding. The future of biotech isn't about making everyone a programmer - it's about creating systems that let scientists focus on science. What's the biggest challenge you've faced in collaborating across the bench-computational divide?

Your assay data is telling you something important. But can you hear it over the noise of: - Spreadsheets scattered across drives - Results buried in slide decks - Analysis redone multiple times by different teams - Hours spent just getting data into the right format At most biotechs, scientists spend more time searching for and managing data than actually analyzing it. The result? Delayed decisions, frustrated teams, and valuable insights left undiscovered. While everyone's science is different, the fact of the matter is that 80% of routine assay analysis follows the same basic patterns. By standardizing data ingestion and analysis, scientists can focus on hard questions, not data wrangling. The biotechs moving fastest aren't just doing more experiments - they're extracting more value from each one. What's holding your team back from faster insights? Comment below or tag someone who's passionate about modernizing drug development workflows.

Yesterday we had our most recent Sphinx Bio Challenge -- Hot Ones Science Bowl Trivia. It was 🌶️🌶️🌶️ (2m+ Scoville) to say the least. Congrats to our newest holder of the Sphinx, Alden!

Excited to announce we're hiring for a Senior Software Engineer! We've seen a lot of interest in what we've been building over the past few months and we'd like to bring on an experienced engineer to help scale our systems. If you're interested in helping improve human health and fix the climate, like LLMs, and want to work on hard problems -- please reach out! It's an exciting time for us at Sphinx Bio and we're looking forward to bringing you on to helps us to build the next generation of biotech infrastructure. See more details here: https://lnkd.in/gzsfZznV

Want to wipe $12B off your market cap? Excel strikes again Excel is good at many things, but its ubiquity in biotech and pharma is extremely dangerous https://lnkd.in/dyDXvT43

Excited to be talking with Jesse Johnson about using AI to unlock value for existing data for biotechs! There's a ton of excitement about AI in bio, but there hasn't been as much discussion of tactics and how you can use what's available today to accelerate your research processes. We’re calling it “Unlocking Biotech Data: How Sphinx leverages AI to help biotechs move faster and make better use of existing data.” It’ll be on Thursday November 21st at 2pm EST/11am PST. You can sign up here: https://lnkd.in/gTXXgJg4

What's the most frustrating routine task eating up your team's time? One thing I see over and over again is brilliant computational biologists running around putting out fires and doing ad-hoc analysis instead of diving into deeper questions. Why? Because the data "needs to be analyzed right away." But here's the uncomfortable truth: 80% of biochemical assay analysis is completely routine. The same calculations. The same graphs. The same statistical tests. Over and over. This isn't a data science problem. It's a tooling problem. Your data scientists should be: - Building ML models to predict drug candidates - Analyzing complex NGS datasets - Creating novel computational methods Not writing scripts to parse platemaps in Excel. The solution isn't hiring more data scientists. It's giving bench scientists the right tools to handle routine analysis themselves. We built Sphinx Bio because we believe both teams deserve better. Bench scientists get automated analysis they can trust. Data scientists get to work on actually challenging problems. Everyone wins and drugs get into the clinic faster.

🎉 It's that time of the year again -- our 2024 State of Techbio Survey is now live! Survey will close on Sunday December 1. Take the survey here: https://lnkd.in/dFJ7H4nr This survey is an opportunity for you to help shape the understanding of the use of software in the life sciences. Its goal is to form a shared understanding of trends, progression, and the general state of tech & life sciences year over year. The survey is anonymous and data/insights will be shared publicly in early 2025. You can find last year's results here: https://lnkd.in/gwy2VZ8W The survey will take about 5-7 minutes to complete. Would really appreciate you all helping amplify with your companies/networks! Thank you in advance

"Our lab is fully automated, so our data is completely consistent." I hear this all the time, and while I would love it to be true, it's one of the most dangerous myths in modern biotech. While automation CAN make your data much more consistent, it can also create a false sense of security. Think about it: - Reagent lots still vary - Cells still behave differently day-to-day - Equipment still drifts - Edge cases still happen It takes a lot of hard work to ensure consistency in the lab. You’ll usually only be able to identify the harder edge cases once you’ve got your data into the right spot. If you’re analyzing each experiment as a one-off, you’ll be unable to see the larger picture and catch these inconsistencies. However, the answer isn’t to just have a set of inflexible analysis scripts. As soon as your teams’ hypotheses change (and they will change), you’ll be scrambling to fix your analysis pipelines. You need a system that helps with standardization while encouraging scientists, automation engineers, and data teams to dig deeper into their data. Striking the balance between flexibility and consistency is always hard in a biotech. Adding in automation just increases the complexity. But if you’re able to do it right, that’s where the magic happens.

AI is not magic. It’s not going to do the science for you (yet). But it’s still incredibly helpful for any sort of digital task. If you’re a scientist, you usually know *what* you want to do — it’s just the *how* that might be difficult or painful. For example, let’s say you just ran a multi-plate concentration response experiment. You know you want to take that data, join it with your platemap, fit a curve to your data, then make some plots. That’s great — you have the beginning of an analysis SOP right here (even if it's just in your head). How might you do that today? Copy and paste from a few Excel sheets, reshape the data, copy into PRISM, fumble around with the settings for a while, then copy the plots into your slides. If it’s a new experiment (or a large one), it might take an hour or two. If you’ve done this before, maybe it only takes half an hour or less. But that’s a complete waste of your time. You don’t need a PhD to copy and paste data between Excel sheets. That’s where the magic of AI comes in — it takes your *what* and solves the *how*. So the next time you’re upset that “AI” isn’t solving all your problems, make sure you’re using it to solve the how, not the what.