Best Practices: Reading the Field Well
This chapter distils the habits that keep biotech literacy alive: following the right sources, recognising hype patterns, and knowing what to update when.
The Biotech News Problem
Biotech news has characteristic failure modes:
- Every advance is "breakthrough": most are incremental
- Every Phase 1 result is "promising": most don't translate
- Every mouse-model cure is imminent: almost all fail in humans
- Every AI drug-design paper changes everything: some change something
A reader without calibration gets whipsawed. Every month appears to bring a new miracle cure; few materialise. Without context, this either produces cynicism ("biotech never delivers") or naïve excitement ("everything is about to change").
Calibration is the skill.
Hype Indicators
Red flags that a story is more hype than substance:
"Could lead to" language
"Could lead to a cure for X" usually means "we found something interesting in a mouse / in a cell line / in a preliminary trial, and years of work would be needed for it to matter in humans".
Not dismissable. Calibrate the confidence to the distance from applicability.
Extrapolating from animal models
Works-in-mice-works-in-humans is roughly a 10% rate in oncology and similar in other fields. Most mouse cures don't translate.
When you read "a new drug cured X in mice", the honest translation is "an interesting signal that will now need many years of work to see if it applies to people". Still worth covering; read accordingly.
Pre-clinical stage reported as clinical
A company claims their drug is "in development" or "shows promise". If they haven't started Phase 1 trials, the data is from cells or animals. Much lower confidence than clinical data.
Single-arm trials
Trials without a control group. A 40% response rate sounds great until you learn that 30% respond to placebo or standard of care. Always ask "compared to what?".
Biomarker outcomes
A drug "reduces biomarker X" is a step removed from "patients live longer" or "patients feel better". Biomarker endpoints are useful but have a history of not correlating with patient outcomes.
Rare-disease specificity dressed as general
A drug that works for 50 patients with a specific rare genetic variant is important for those patients. It is often presented as "a cure for Y", where Y is a much broader disease. Read specifically.
Press releases vs peer-reviewed work
Press releases come first, in flattering terms. Peer-reviewed papers come later with more caveats. Always check whether the news is based on the full paper or the company's summary.
AI and biotech mashups
"AI will cure cancer" articles have been appearing for a decade. Some AI contributions to biology are real and important (AlphaFold). Others are marketing. Use standard hype filters: are they showing real outcomes, or just a shiny demo?
Legitimate Progress Signals
Signals that something real is happening:
Phase 3 success
Large, well-controlled Phase 3 trials showing statistically significant benefit on hard endpoints (mortality, progression, symptom scores) are the strongest signal short of approval.
Convergent evidence
Multiple studies, from multiple groups, using multiple methods, pointing the same direction. The HER2-in-breast-cancer story (multiple lines of evidence), CRISPR's development (many labs independently), CAR-T (progressive improvement across trials and diseases).
Replicable mouse models in multiple labs
A single lab's mouse result is weak; five labs replicating it is stronger.
Gradual improvement at known difficult problems
Steady improvement in protein-structure prediction, from modest to excellent over 20 years, is a healthier signal than sudden "breakthroughs" in areas with no prior signal.
Real-world data after approval
Once drugs are in routine use, their real effect sizes become clearer. Some approved drugs turn out to help less than the trials suggested; others hold up. Long-term real-world data is a good reality check.
Sources Worth Following
A mixed diet:
Peer-reviewed literature
The primary source. Journals that matter:
- Nature, Science: general science, high-profile
- Cell: cell and molecular biology
- The New England Journal of Medicine (NEJM): major clinical trials
- Nature Biotechnology, Cell Stem Cell: topical
- PNAS: US National Academy of Sciences
PubMed.gov is the free database. Many papers are paywalled; preprint servers (bioRxiv, medRxiv) have preprints.
Trade press
- STAT News: biotech-focused, good clinical coverage
- Endpoints News: industry deals and trials
- FierceBiotech: industry news
- Biopharma Dive: clinical and regulatory
Blogs and newsletters
- In the Pipeline (Derek Lowe, Science blog): pharma chemistry, sceptical, essential
- Scott Alexander's ACX: occasional deep dives into biomedical topics
- Lab Notes, The Transmitter, etc.: varied quality; sample
Longer-form
- Nature News & Views, Science Perspectives: commentary on new findings
- Podcasts: This Week in Virology, Ground Truths (Eric Topol), The Drive (Peter Attia) for clinical
- Books (see the tutorial's README resource list)
What to skip
- Mainstream press headlines on specific drugs (often misleading without the underlying paper)
- LinkedIn thought-leadership on biotech (mostly marketing)
- Twitter threads from clear partisans without specific claims
Reading a Paper
A minimal approach for non-scientists:
- Read the abstract: what they claim in a paragraph
- Look at the main figure(s): the data usually tells the story
- Read the discussion: limits and caveats
- Note the funding and authors: who paid? Who wrote? Conflicts of interest?
- Check replication: has this been replicated, or is it the first report?
A first-report single-lab paper is interesting but tentative. A replicated finding across multiple labs is more solid.
You don't have to understand every methodological detail. You can form useful judgments from the overall shape.
The Replication Problem
Biomedical research has a replication problem. Studies in the mid-2010s showed that a large fraction of high-profile findings fail to replicate in independent labs. Cancer biology, psychology, and some aspects of molecular biology have been particularly affected.
What this means:
- Treat single studies as preliminary until replicated
- Favour areas where replication has been done
- Be especially sceptical of surprising findings that haven't been independently tested
This isn't a reason to dismiss the field. It's a reason for calibrated scepticism about specific claims.
Conflict of Interest
A lot of biotech research is done by people with financial stakes in the outcome: paid consultants, equity holders, founders. This doesn't invalidate their work, but it's worth knowing.
Standards vary:
- Some journals require rigorous disclosure
- Some don't
- Some conflicts are disclosed clearly; some are buried
Worth reading the conflict-of-interest statements when a claim is surprising or specific to a company's product.
Picking a Focus
Biotech is enormous. You'll get more from narrowing:
- Pick one disease area (cancer, neurodegeneration, infectious, metabolic, rare)
- Pick one technology (CRISPR, mRNA, cell therapy, AI drug design)
- Pick one company and follow its pipeline
- Pick one academic lab whose work interests you
Deep in one area produces more useful understanding than shallow across everything.
Keeping Up Without Burning Out
Biotech never stops moving. A sustainable approach:
- One newsletter, weekly: keeps you current without consuming you
- One long-form piece per month: deeper context on something that's happened
- One paper per month or quarter: actual primary literature on something specific
- Quarterly catch-ups on your chosen focus area
Don't try to read everything. Most of it won't matter in five years. The 10% that will becomes clearer with patience.
Calibrating Predictions
A habit worth cultivating: make predictions (even to yourself), write them down, check them later.
- "I think this Phase 2 will read out positively"
- "This drug will be approved within 3 years"
- "This platform will not deliver on current promises"
Most predictions are wrong. But tracking your own lets you calibrate. After a year, you'll see whether you're optimistic, pessimistic, or uncalibrated in specific ways.
Well-calibrated people are rare and valuable. The field rewards calibration because it makes investment, hiring, and research decisions better.
The Long Arc
Biotech moves in decades. Individual stories resolve in months; the big themes take twenty years.
Decades-long arcs visible now:
- Sequencing getting cheap (1990s-present): still unfolding
- Cancer moving from generic poisons to targeted therapies (1990s-present)
- Immune-system-based cancer treatment (checkpoint inhibitors, CAR-T) (2010s-present)
- Gene editing becoming routine (2012-present)
- Molecular understanding of neurodegeneration (ongoing; slow)
To read biotech well, hold both timescales: this quarter's news and this decade's direction. Neither replaces the other.
The Limits of Literacy
A closing note: this tutorial makes you biotech-literate, not biotech-competent. You can follow the news; you can't design a drug. That's fine. Literacy is valuable in its own right.
If you want to go deeper, you'll need specialisation: a graduate program, a serious bootcamp, or years of self-directed work. The literacy you've built is the foundation for any of those paths.
If literacy is enough for your purposes (reading, investing, writing, voting), you have what you came for. Maintain it by reading, updating, and occasionally dropping a chapter of a textbook.
Where to Go From Here
- Pick one disease area or technology and follow it for a year
- Subscribe to STAT News, Endpoints, or In the Pipeline
- Read Siddhartha Mukherjee's The Gene if you haven't
- Follow one clinical trial you find interesting from Phase 1 through whatever its outcome
- Revisit this tutorial in a year; fields move, chapters will age, and new ones will matter
Biology is, by a long margin, one of the most interesting intellectual territories available. The tutorial is over; the interest doesn't have to be.