Building Block Libraries: Buy, Build, or Commission?

Building block libraries in drug discovery get discussed as a procurement line item and decided as a strategy question, which is why so many of them disappoint. A medicinal chemistry lead with a validated hit and a budget faces a choice that looks like three prices on three quotes but is actually three different bets on time, intellectual property, and where the team’s attention goes for the next two quarters.

The framing that makes this decision tractable is to separate design from sourcing. The design question — which compounds do I want and why — is chemistry. The sourcing question — buy, build, or commission — is operations. Teams that answer them in that order get useful structure-activity relationship data. Teams that answer them in the opposite order end up with a freezer full of compounds that all look alike.

This post covers both, in that order: what makes a library well-designed, then the honest economics of the three routes to material, then the execution detail that decides whether a commissioned library actually arrives usable.

Library Design Beats Library Size

A well-designed 500-compound set explores more useful chemical space than a poorly-designed 5,000-compound set. This is the single most important idea in library work, and it is counterintuitive to anyone whose instinct is that more screening data is better data.

The reason is redundancy. Compounds cluster. A large commercial set assembled by “what was cheap and available” will contain hundreds of close analogs of the same handful of privileged fragments, because those are the fragments that are cheap and available. You pay for 5,000 wells and receive maybe 400 genuinely distinct structure-activity questions. A designed 500-compound set, chosen to spread across a defined property window with deliberate coverage of shape and substitution pattern, asks close to 500 distinct questions.

The Cost of Redundancy Is Not Just Money

Redundant compounds cost assay capacity, which is usually more constrained than compound budget. They also cost interpretive clarity. When forty near-identical analogs all come back at the same potency, you have learned one thing forty times over and consumed forty wells doing it. Worse, a cluster of actives that are all variations on one fragment can look like a robust structure-activity trend when it is a single data point wearing a costume.

Design Coverage Is Measurable Before You Buy Anything

You do not need the compounds to know whether the set is well spread. Compute the descriptors, plot them, and look at the holes. A design review that happens before purchase orders go out costs an afternoon and routinely changes which third of the list gets bought. The broader case for designed rather than acquired libraries is covered in our overview of combinatorial chemistry libraries in drug discovery.

The Design Principles That Matter

Good library design comes down to five checks. Run all five before committing to a compound list.

Property Space Coverage

Compute and plot the standard descriptor set across your candidate list, then look at the distribution rather than the mean:

Descriptor	What it controls	Typical lead-like window
Molecular weight (MW)	Room to grow during optimization	200 to 350 Da for the analog set
Calculated logP (cLogP)	Lipophilicity, solubility, promiscuity	1 to 3 for the analog set
Topological polar surface area (TPSA)	Permeability and efflux liability	40 to 90 square angstroms
Rotatable bonds	Conformational entropy cost of binding	3 to 7
Hydrogen bond donors (HBD)	Permeability, crystallinity	0 to 2
Hydrogen bond acceptors (HBA)	Solubility, binding interactions	2 to 6

The point is coverage, not compliance. If every compound in your set sits at cLogP 3.5, you cannot learn whether lipophilicity is driving your potency or your assay artifact. Aim to populate the corners of the window as well as the middle.

Lead-like, not drug-like. Analog sets for early structure-activity work should sit below the Lipinski ceilings, not at them, because optimization adds molecular weight and lipophilicity. A hit that already sits at 480 Daltons and cLogP 4.5 has nowhere to go. Design the set 100 to 150 Daltons lighter than where you expect the final molecule to land.

Lipinski and Veber, and When to Break Them Deliberately

The Lipinski rule of five (molecular weight under 500, cLogP under 5, hydrogen bond donors under 5, acceptors under 10) and the Veber criteria (rotatable bonds at or below 10, polar surface area at or below 140 square angstroms) are descriptive summaries of oral drugs, not laws. They are useful as a default and misleading as a gate.

Break them on purpose when the target class demands it. Macrocycles, protein-protein interaction inhibitors, kinase degraders and certain antibacterials routinely sit outside the rules and still work. What matters is that you decided to leave the window rather than drifted out of it. A library brief that says “deliberately targeting 550 to 700 Daltons with polar surface area above 140 because the target is a protein-protein interface” is a design. A library that ends up there because the fragments were convenient is an accident.

Three-Dimensionality and the Escape From Flatland

The fraction of sp3-hybridized carbons, written Fsp3, is the standard shorthand for how three-dimensional a molecule is: the count of sp3 carbons divided by total carbon count. Flat, aromatic-rich compounds have low Fsp3. The “escape from flatland” argument is that libraries built predominantly from Suzuki couplings of aromatic partners drift toward low Fsp3, and that these compounds tend to show worse solubility, weaker selectivity and higher attrition than more saturated counterparts.

Practically, that argues for deliberately seeding your diversity set with saturated and semi-saturated partners. Small saturated heterocycles are the workhorses here. Azetidines such as 1-BOC-3-Iodoazetidine (CAS 254454-54-1) and 1-BOC-3-Aminoazetidine (CAS 193269-78-2) give a compact, high-Fsp3 vector with a protected nitrogen that unmasks cleanly for a second round of diversification. Fluorinated saturated amines like 4,4-Difluoropiperidine (CAS 21987-29-1) shift basicity down while holding shape, which is often exactly what a compound with an unwanted hERG or phospholipidosis signal needs. Chiral saturated amines such as (S)-2-Methylpyrrolidine (CAS 59335-84-1) add a stereocenter, which doubles the structure-activity question at essentially no synthetic cost when the enantiopure material is available. The piperidines and piperazines category and the BOC-protected compounds category are the two shelves most discovery teams pull from for this purpose.

Ring systems that push Fsp3 harder — spirocycles and bridged bicyclics — carry a synthesis cost that is worth understanding before you design them in. We cover the tractability question in spirocyclic and bicyclic amine synthesis CRO capabilities.

Synthetic Tractability of the Exit Vectors

An exit vector is the bond from your scaffold at which a diversity element attaches. Every library design implicitly commits to a set of them, and this is where designs most often fail on contact with reality.

Check three things per vector:

1. Is the vector accessible on the scaffold at usable scale? A vector that requires a five-step protecting group dance to expose is not a library vector, it is a separate program.
2. Is the coupling chemistry robust across the whole partner set? One reaction that works for 80 percent of your amines and fails for the hindered and heteroaryl ones will systematically bias your data toward small, unhindered substituents.
3. Does the vector point where the structure-activity data says it should? If your co-crystal or docking model says the growth direction is solvent-exposed at position 4, a library that diversifies position 2 will be well-executed and uninformative.

Functional Group Compatibility With Your Coupling Chemistry

This is the check that saves the most money and gets skipped the most often. Run your candidate building block list against the reaction you plan to use and flag the incompatibilities before you buy.

• Amide couplings tolerate a great deal but will acylate free secondary amines elsewhere in the partner. Flag any building block with a second unprotected nitrogen.
• Suzuki–Miyaura couplings are sensitive to free carboxylic acids competing for the palladium, to certain unprotected heterocyclic nitrogens that poison the catalyst, and to protodeboronation on electron-poor and 2-heteroaryl boronic acids. A partner such as 4-Bromophenylboronic acid (CAS 5467-74-3) is useful precisely because it is bifunctional, but that same bifunctionality means it will oligomerize if your conditions are not chemoselective.
• Buchwald–Hartwig aminations are ligand-dependent in a way that is partner-specific. Primary aliphatic amines, anilines and secondary cyclic amines often want different ligands.
• Nucleophilic aromatic substitution on activated heteroaryl halides — the workhorse for scaffolds like 2-Chloropyrimidine (CAS 1722-12-9) — is beautifully tolerant but rate-sensitive to amine nucleophilicity, which means your least basic partners may simply not react under plate conditions.
• N-alkylation and N-arylation of azoles carries a regiochemistry problem. A substrate like 3-Bromo-1H-indazole (CAS 40598-94-5) has two nitrogens, and the N1 versus N2 ratio depends on base, solvent and electrophile. If you do not resolve and assign the regiochemistry, half your structure-activity data describes a compound you did not intend to make.

Sourcing across these classes is straightforward from a broad catalog — the boronic acids and heterocyclic compounds categories cover most of the common coupling partners, and the full product catalog runs to over 7,000 compounds — but breadth of catalog is not the same as fitness for your specific reaction. Screen for compatibility, not availability.

Option 1: Buy From Catalog

Buying from catalog is the fastest and cheapest route per compound, and the right default for early structure-activity breadth and for tool compounds.

Material ships in days rather than months. Unit economics are as good as they will ever get, because you are buying compounds that were made once and sold many times. There is no synthesis risk, no development time, and no draw on your own laboratory.

The Real Weakness Is Not Quality, It Is Universality

Catalog compounds are available to every one of your competitors on the same terms. That has two consequences. First, your intellectual property position on any composition-of-matter claim built purely from catalog fragments is weaker, because the compounds themselves are prior art or trivially accessible; the novelty has to live in the scaffold combination rather than the pieces. Second, well-known catalog fragments have often already been screened against your target class by somebody, which means the obvious hits may already be claimed.

That is a reason to be thoughtful about catalog use, not to avoid it. For establishing breadth of structure-activity relationship at the start of a program, for building the negative-control and selectivity panel, and for tool compounds where novelty is irrelevant, catalog is simply the correct answer.

How to Select Well From a Large Catalog

The failure mode is buying what is cheap rather than what is informative. Selecting well takes a few hours of computational work:

1. Filter hard on functional group compatibility first. Remove anything that will not survive your coupling chemistry. This typically eliminates 20 to 40 percent of a naive list and costs nothing.
2. Apply structural alerts. Remove pan-assay interference compounds, Michael acceptors, aggregators and known frequent hitters unless you have a specific reason to keep them.
3. Cluster and select representatives. Compute fingerprints, cluster at a similarity threshold, and take a small number of representatives per cluster rather than everything in the cluster. This is where the redundancy comes out.
4. Fill property holes deliberately. Plot the surviving set in property space and buy specifically to fill the gaps, even when those compounds cost more per gram.
5. Check availability at your real scale. A compound listed at 10 milligrams with a six-week lead time is not a library member, it is a promise.

Step three is the one that separates a designed catalog purchase from a shopping list, and it is the reason a smaller, better-selected order routinely outperforms a larger, cheaper one.

Option 2: Build In House

Building in house gives full control over design and unambiguous intellectual property ownership, at the cost of your chemists’ time, which is nearly always the scarcest resource in a discovery organization.

The economics are frequently misread because internal labor is treated as free. It is not. A medicinal chemist spending six weeks producing forty analogs is a medicinal chemist not spending six weeks on route design, on interpreting the last data set, or on the next scaffold. Fully loaded, that time is expensive, and it is expensive in the currency you have least of.

When Building In House Is Right

• The chemistry is core to your program. If the key transformation is novel, is itself patentable, or requires judgment that cannot be handed over in a specification, keep it inside.
• The compounds are strategically important. Late-stage analogs around a lead you intend to file on belong in your own hands and your own notebooks.
• The scale is small and the iteration loop is tight. Making six compounds in response to yesterday’s assay result is faster in house than through any external process, every time.
• The material is hazardous, controlled, or otherwise awkward to ship. Some chemistry does not travel.

When It Is the Wrong Answer

Analog production is the wrong use of a senior medicinal chemist when the chemistry is known, the transformation is a single robust step, and the only variable is which partner goes in which well. That work is parallelizable, and parallelizing it inside a discovery lab usually means one chemist doing it serially. The tradeoff between internal and external chemistry capacity is worked through in more detail in contract R&D versus in-house labs: a cost analysis.

Option 3: Commission a Custom Library From a CRO

Commissioning is the middle path: you own the design and the intellectual property, someone else runs the parallel synthesis and purification.

The structure of the arrangement matters. In a well-run commissioned library, the client specifies the scaffold, the diversity element list, the purity and quantity specification, and the plate format. The contract laboratory executes, purifies, characterizes and ships. The design and the resulting compounds belong to the client under the contract. This is different from buying a vendor’s pre-existing “targeted library”, where you are back in catalog territory with the same universality problem.

When Commissioning Is the Right Call

The clearest case is a focused set around a validated hit, when your internal chemists should be doing route work rather than analog production. You have a confirmed hit, a structure-activity hypothesis, and a need for 100 to 400 compounds around it in the next three months. That is exactly the shape of work that parallelizes well externally and blocks a lab badly internally.

The second good case is chemistry your team can specify but does not run routinely — a fluorination, a high-pressure hydrogenation, a chiral separation — where the equipment matters more than the insight.

The Three-Way Comparison

Dimension	Buy from catalog	Build in house	Commission from CRO
Cost per compound	Lowest	Highest fully loaded	Middle
Time to material	Days to 2 weeks	Weeks to months, queued behind other work	6 to 16 weeks typical for a focused set
Intellectual property position	Weakest; compounds are public	Strongest; fully owned and documented internally	Strong; design and compounds owned by client under contract
Design control	Limited to what exists	Complete	Complete, if the specification is written well
Internal resource cost	Near zero	Very high; consumes bench chemists	Low; consumes project management and review time
Scalability of the winner	Depends on the vendor, often poor	Good; the route is already yours	Good, if resupply was scoped at the start

Most programs should not pick one. The common working pattern is catalog for early breadth, a commissioned set for focused structure-activity exploration once a hit is validated, and in-house effort concentrated on route development and the late-stage compounds that matter most. The timeline that surrounds this decision is laid out in hit-to-lead chemistry and the contract R&D timeline.

Executing a Commissioned Library Well

A commissioned library succeeds or fails on six execution details. Get these into the statement of work before signing.

Parallel Synthesis and the Scaffold-Plus-Diversity Strategy

The standard architecture is a common scaffold prepared once at scale, then split across a plate and coupled with a diversity element in each well. This concentrates all the difficult chemistry into a single batch of scaffold, made and characterized properly, and leaves the plate work to a single robust transformation.

The practical implications are worth stating plainly. Make more scaffold than the arithmetic says — 1.5 to 2 times, because you will lose some to failed wells, analytical samples and the inevitable repeat plate. Validate the coupling on three to five representative partners at single-well scale before committing the full plate; picking a hindered one, an electron-poor one and a heteroaryl one for that validation is worth more than picking three easy ones. And run the whole plate under one set of conditions rather than optimizing per well, because per-well optimization is just serial synthesis with extra steps. The broader case for this approach is in combinatorial chemistry accelerating drug discovery.

Purification Approach

Mass-directed preparative HPLC is the default for library purification: the instrument triggers fraction collection on the expected molecular ion, which means it collects your product rather than the largest peak. It is robust, it is automatable, and it handles the reality that every well has a different impurity profile.

Supercritical fluid chromatography is the better choice in two situations: separating enantiomers or diastereomers when the library contains stereocenters, and purifying compounds that are unstable or poorly behaved in aqueous or acidic mobile phases. Supercritical fluid chromatography also strips solvent far more easily, which matters when you are drying several hundred fractions.

Ask which method the laboratory intends to use, and ask what happens to wells that fail to purify cleanly. The honest answer is that some are reported as failures rather than shipped at 70 percent purity, and you want that to be the answer.

Realistic Success Rates

Not every well delivers, and any proposal implying otherwise should be questioned. For well-behaved chemistry on a validated scaffold with a screened partner list, 60 to 85 percent of wells returning material at or above specification is a realistic planning assumption. Harder chemistry, sterically demanding partners or sensitive functionality can push that lower. Novel chemistry on an unvalidated scaffold can be much lower on the first plate and much better on the second, once conditions are tuned.

Plan for it. If you need 200 compounds in hand, design 260 wells. Confirm in the contract whether you pay per well attempted or per compound delivered to specification, because those are very different commercial structures and the second one aligns incentives better.

Quality Control Standards

Specify these explicitly rather than accepting a default:

Parameter	Primary screening	Dose-response and selectivity work
LC-MS purity	90 percent or greater by ultraviolet area	95 percent or greater
Identity confirmation	Molecular ion confirmed by mass spectrometry	Mass spectrometry plus proton NMR
Detection method	State the wavelength; consider evaporative light scattering or charged aerosol detection for chromophore-poor compounds	Same, plus quantitative NMR where concentration accuracy matters
Stereochemical purity	Not usually assessed	Chiral HPLC where a stereocenter is present

The wavelength caveat is not pedantic. Ultraviolet area percent at 254 nanometers systematically under-reports impurities that lack a chromophore and over-reports those with a strong one. For any set containing saturated, non-aromatic compounds, ask for evaporative light scattering or charged aerosol detection alongside the ultraviolet trace. Our overview of analytical testing methods for pharmaceutical buyers covers how to specify these methods so that two laboratories report comparable numbers.

Quantity Per Compound

For primary screening, 3 to 10 milligrams per compound is generally sufficient and is the range most parallel synthesis routes deliver comfortably. That supports a dimethyl sulfoxide stock, a primary screen, a confirmation, and a modest dose-response.

Asking for 25 to 50 milligrams across the whole library is a common and expensive mistake. It raises the scale of every well, which raises reagent cost and often reduces success rate, in order to bank material for compounds that will mostly be inactive. The better structure is a small quantity across the whole set plus a contractual option to resupply selected hits at a larger scale, priced up front.

Plate Format and Logistics

Agree the format before synthesis starts. Ninety-six-well is the practical default for compound handling and shipping; 384-well is for assay-ready plates rather than synthesis output. Specify whether you want dry film or solution, the vial or plate type, the barcode and mapping file format, and how the plate map ties to your compound registration system.

Ask for the plate map as a structured file with a defined column for well position, structure as SMILES, batch identifier, measured purity, measured mass and delivered quantity. A library that arrives without a clean registration-ready file costs a chemist a week of manual transcription, and manual transcription introduces errors that surface months later as an unreproducible result.

The Follow-Up Problem Nobody Plans For

When a hit emerges from a library, you need 10 to 100 times more of that specific compound, quickly — and the parallel route very often does not scale.

This is the most predictable failure in library work and it is almost never scoped at the start. The reasons the plate route fails at scale are structural, not incidental:

• Purification does not scale. Mass-directed preparative HPLC is superb for 5 milligrams and untenable for 5 grams. The scale-up route needs a crystallization, a distillation or a plug-filtration endpoint.
• Reagent excesses that are trivial in a well are unacceptable in a flask. Four equivalents of a coupling reagent is nothing at 20 micromoles and a cost and disposal problem at 200 millimoles.
• Plate solvents and concentrations are chosen for automation, not for throughput. Dilute dimethyl sulfoxide or N-methylpyrrolidone conditions that work fine in a well become a solvent volume and workup problem at scale.
• Exotherms and mixing appear only at scale. A reaction that is invisibly exothermic in 200 microliters is a real thermal event in a 2-liter flask.
• The diversity element may not be available in quantity. This is the most common one. The building block you bought at 100 milligrams for the plate may have a twelve-week lead time at 100 grams, or may not exist at that scale.

The fix is to plan the resupply route at design time. Three concrete actions, all cheap if done early:

1. Check bulk availability of every diversity element before you commit the plate. If a building block cannot be had at 100 grams within your program timeline, decide consciously whether you still want the analog. Most catalog compounds have a scale cliff somewhere.
2. Write the resupply option into the contract. Agree pricing and lead time for 1 to 10 gram resupply of selected wells at the time you sign, not at the moment you need it.
3. Ask the laboratory to sketch a scalable route for the scaffold in parallel with plate execution. This is inexpensive as a paper exercise and saves months when a hit lands. The considerations are covered in scale-up challenges in custom synthesis and process chemistry optimization from lab to pilot.

If your library contains stereocenters, the resupply question has a second layer: whether to resupply by resolution or by asymmetric synthesis. That decision changes with scale and is worked through in chiral resolution versus asymmetric synthesis.

Storage, Handling, and Compound Integrity Over Time

Library compounds degrade in storage, and degraded compounds produce results that look like biology.

Dimethyl Sulfoxide Stocks

Ten millimolar in dimethyl sulfoxide is the standard, stored at minus 20 degrees Celsius in sealed, low-moisture conditions. Dimethyl sulfoxide is strongly hygroscopic — an open plate on a bench absorbs water measurably within hours, and accumulated water drives both precipitation of poorly soluble compounds and hydrolysis of susceptible ones. Keep plates sealed, keep them dry, and let them equilibrate to room temperature before opening so that condensation forms on the outside of the seal rather than inside the well.

Freeze-Thaw Degradation

Repeated freeze-thaw cycling is the largest controllable cause of library decay. Each cycle admits moisture and drives some compounds out of solution, and material that has precipitated does not fully redissolve on the next thaw, so the effective concentration drifts downward silently. The mitigation is straightforward: aliquot into single-use daughter plates on the first thaw and never cycle the mother plate more than a handful of times.

Compound Integrity Over Time and Periodic Re-Checks

Different chemotypes decay at very different rates. Esters, activated halides, aldehydes, thiols and Michael acceptors are the usual first casualties; robust heteroaromatic amides can sit for years. Because the rate is chemotype-dependent, a blanket expiry date is the wrong tool.

A workable regime is a representative sample re-checked by LC-MS annually, with a targeted check on any compound before it drives a program decision. A cheap identity-and-purity re-check ahead of a decision-grade experiment is far cheaper than chasing an irreproducible result for a quarter. Setting the specification and the re-check method so results are comparable over time is covered in reference standard sourcing and certification.

For fluorescent or chromogenic reporters in a screening set, photostability is a separate axis of decay that ordinary purity checks miss entirely; that problem is treated in sourcing specialty dyes and fluorescent probes.

Putting the Framework Together

The decision sequence that works, in order:

1. Define the biological question. Which vector, which property window, how many distinct structure-activity questions do you actually need answered?
2. Design the set. Property coverage, three-dimensionality, exit vectors, functional group compatibility. Review the plots before anyone raises a purchase order.
3. Split the list by route. Catalog for what exists and is compatible. Commission for what does not exist but is routine chemistry. Keep in house only what is genuinely core.
4. Specify quality and quantity honestly. Small quantity, clear purity threshold, realistic well success rate, registration-ready data file.
5. Scope resupply before you start. Bulk availability of building blocks, contractual resupply pricing, a paper route to scale for the scaffold.
6. Store it properly and re-check it. Sealed dimethyl sulfoxide stocks, minimal freeze-thaw, periodic LC-MS.

Teams that follow this sequence spend less and learn more, mostly because they stop paying for compounds that answer questions they were never asking.

For teams weighing where to place the work geographically, our US-based custom synthesis capability map covers who does what domestically. External references worth having on hand: the National Center for Advancing Translational Sciences assay guidance manual for screening and compound quality practice, the American Chemical Society Journal of Medicinal Chemistry author guidelines for the compound purity and characterization standards reviewers expect, and the NIST Chemistry WebBook for physical property reference data.

ChemContract Research runs US-based combinatorial and automated synthesis for commissioned library work, custom synthesis from milligram to multi-ton for scaffold preparation and hit resupply, and contract R&D for route development when the parallel route will not scale. Our catalog covers over 7,000 building blocks, and our analytical group runs LC-MS, NMR and chiral HPLC in house for library characterization. Send us your scaffold and diversity element list and we will return a design review, a realistic well-count estimate and a quote within 24 hours — start here.

Frequently Asked Questions

What is a building block library in drug discovery?

It is a curated collection of small, synthetically reactive fragments — amines, boronic acids, halides, carboxylic acids and similar — chosen so they can be coupled onto a common scaffold to generate analogs for structure-activity relationship work. The library is the set of diversity elements, not the finished compounds, though the term is often used loosely for both.

Is a bigger library always better?

No. Coverage of property space and compatibility with your coupling chemistry matter far more than raw count. A 500-compound set spread deliberately across molecular weight, lipophilicity, polar surface area and three-dimensionality gives more usable structure-activity information than 5,000 compounds clustered in one flat, greasy corner of chemical space.

How much does a commissioned custom library cost per compound?

It varies widely with scaffold complexity, number of synthetic steps per analog, and purity specification. Simple one-step parallel couplings at 10 to 50 milligram scale with mass-directed purification sit at the low end; multi-step sequences with chiral separation sit far higher. Ask for a per-well quote that states the assumed step count and the purity threshold.

What success rate should I expect from parallel synthesis?

Not every well delivers. For well-behaved chemistry on a validated scaffold, 60 to 85 percent of wells returning material at or above the purity specification is a realistic planning assumption. Harder chemistry, hindered partners, or sensitive functionality can drop that materially. Design the plate with the expected dropout in mind.

What purity standard should a screening library meet?

Ninety percent by LC-MS ultraviolet area with a confirmed molecular ion is a common primary-screening threshold, with 95 percent typical for compounds moving into dose-response or selectivity work. Specify the detection wavelength and whether evaporative light scattering or charged aerosol detection is also required, because area percent varies by detector.

How should screening compounds be stored?

As dimethyl sulfoxide stock solutions at 10 millimolar in sealed, low-moisture conditions at minus 20 degrees Celsius, with freeze-thaw cycles minimized. Dimethyl sulfoxide is hygroscopic and water uptake drives precipitation and hydrolysis. Re-check integrity by LC-MS periodically, particularly before committing a compound to a decision-grade experiment.