Skip to content
Forticia
Research
Quantitative FinanceEquities, FX, futures. Factors and backtests, every run replayable.Computational BiologySequence, folding, simulation. Versioned, reproducible labs.Cultural IntelligencePhilosophy, governance, ethics. How institutions decide and answer for it.AI InstrumentationPrivate models. Multi-agent orchestration. Guardrails on write.View all research
Quantitative Finance
  • Equities, FX, futures
  • Factors and backtests
  • Every run logged and replayable
InfrastructurePapersPolarisLink™About
Sign inRequest access
Request access

Research

Quantitative FinanceEquities, FX, futures. Factors and backtests, every run replayable.Computational BiologySequence, folding, simulation. Versioned, reproducible labs.Cultural IntelligencePhilosophy, governance, ethics. How institutions decide and answer for it.AI InstrumentationPrivate models. Multi-agent orchestration. Guardrails on write.

Platform

InfrastructurePapersPolarisLink™About
Request accessSign in
Forticia

A private institute for computational research. It publishes original research and runs a governed environment where every run is logged and replayable.

Sign inSystem status

Research

Quantitative FinanceComputational BiologyCultural IntelligenceAI Instrumentation

Infrastructure

WorkspacesHelper planeGoverned AIPapersStatus

PolarisLink™

SpecificationQuickstartClientsChangelogGitHub repository

Institute

AboutFAQRequest accessContactSign inForticia on GitHub

Forticia publishes research and simulations. Nothing on this site is investment advice.

Official channels: forticia.uk and github.com/Forticia

© 2026 ForticiaPrivacyTerms
Nresidue 1
H1
H2
S1
S2
H3
Cresidue 89
  1. Home
  2. Research
  3. Computational Biology

Computational biology research at Forticia: Sequence, structure,
and the run that
made it.

Genomic sequence analysis, protein folding visualisation and simulation frameworks, held so a result can be opened again. Same inputs. Same parameters. Same environment.

Request accessHow a run is recorded
Channels
Four, mediated
Reruns
Deterministic where the tools allow
Writes
Lead-approved

Local order

Stretches of the chain settle first.

Short runs of residues adopt repeating local shapes, helices and strands, joined by flexible loops. Colour now follows that assignment. Nothing is packed yet: this is the shape of the parts, not of the whole.

  • Helix
  • Strand
  • Loop

Packing

Then the chain closes on itself.

Residues far apart in sequence end up close in space. Each thin line joins a pair that sits near the other in the packed chain. That contact pattern is what predicted structures are compared on, and what a simulation has to keep intact.

Inspection

Read it one residue at a time.

Drag to turn the chain. Hover a residue, here or in the sequence below, to see its side chain and contacts. Change the representation from the rail on the right and what the colour encodes from the key beneath. Every number in the status line is computed from the coordinates on screen.

HelixStrandLoop
Secondary structure colouring
Chain A89 residuesHover to inspect
  1. Sequence
  2. Local order
  3. Packing
  4. Inspection
MGKSTLQLALEQLKEAKKLLASNGSGLLKAAEALLAAEMLLDKDNDYVTIYVWNGGTVIVIIWFNGSPLEQAEEKLARAALKLELDSGG

State of the field

What is routine now,
and where it still breaks.

Some problems moved from research to routine in a few years. The edges are where results get over-claimed, so that is where the checks go.

AreaRoutineStill hardAsked of a result here

Sequence analysis

HMMER, HH-suite, MMseqs2

Profile-HMM and fast k-mer search tools scale homology search to very large databases. Clustering by identity is a standard first step.

Remote homology below detectable similarity. Protein language model embeddings are being used for search and compare well in published studies, but they inherit what the model has seen.

Database release, aligner version and sensitivity settings recorded with every search.

Structure prediction

AlphaFold 3, Boltz-1, CASP

Single-domain folds. The CASP16 monomer assessment described single-domain fold prediction as nearly solved.

Ranking complexes, stoichiometry, change on binding. Diffusion-based predictors can draw spurious order in disordered regions. A prediction is one static state, and more seeds do not recover an ensemble.

Confidence kept beside the coordinates, with disordered regions marked as such.

Protein design

RFdiffusion, ProteinMPNN

Generate a backbone, assign a sequence, filter in silico on self-consistency and predicted error.

Reported experimental success rates sit near one in five on curated targets and are far lower in some independent applications. The filter is not the assay.

Filter thresholds and every rejected design logged, so a hit rate has a denominator.

Simulation

Classical MD, metadynamics, machine-learned potentials

Classical force-field dynamics with enhanced sampling. Machine-learned potentials are narrowing the gap to quantum accuracy.

Femtosecond timesteps against events that take microseconds to days. Sampling, not compute, is usually what limits a conclusion.

Topology, timestep, thermostat and seed travel with the trajectory. A run that cannot restart is incomplete.

Single-cell and omics

Community best-practice workflows, Harmony

Quality control, normalisation, variable-gene selection, embedding and clustering follow community best-practice guides.

Separating biology from technical variation. Integration methods can over-correct or add artefacts, and benchmarks do not always agree.

Batch structure declared up front. Integration tuned on one split and scored on another.

Evaluation

MMseqs2 clustering, time and family splits

Held-out test sets and cross-validation.

Random splits leak through homology. Pretrained models can leak through their corpus: one study scored higher on a split that ignored pretraining clusters than on a harder pretraining-aware one.

Cluster-aware, pretraining-aware splits, with the split itself part of the record.

Sequence analysis

Conservation is the evidence
in a sequence alignment.

Homology search lines related sequences up column by column. What stays fixed across the family is what the family cannot afford to change, and that is the signal structure and function predictions lean on. Conservation, information content and occupancy below are computed from the rows on screen.

A constructed family of 14 sequences derived from the chain in the viewer above with a simple substitution model, in which buried positions change least. It stands in for the real alignment a search would return, and the tracks are computed the same way.

Colour
Order
Position
1020304050607080
Structure
Conservation
Information (bits)
GKSTLQLALEQLKEARKKLLASNGSGLLKAAEALLAAEMLLEDKREDNDYVTIYVWNGGTVIVIIWFNGSPLEQAEEKLARAALKLELEDSGG
Occupancy
queryreference
MGKSTLQLALEQLKEAKKLLASNGSGLLKAAEALLAAEMLLDKDNDYVTIYVWNGGTVIVIIWFNGSPLEQAEEKLARAALKLELDSGG
homolog_0189%
GGKSTLQLALEQLKDAKKLLASQGSGLLKAAEALLAAEMLLDKDNDYVQIYVWNPGTVIVIIWFNG–PLEQAEEKLARAALKIEIESGT
homolog_0291%
MGKSTLQLALEQLKEAKKLLASNGSGLLKVAEALLAAEILLE–DNEYVCIYVWNGPTVIVIIWFNGSPLEQAEEKLARAALRLEIDSGG
homolog_0378%
VWKVTMNLATDQLKEAKKLLASL–––LLRAAEALLAAEVLCDKENDYATIYVWNPLTVIMIIWFNGSPLEQAEEKLARAVLKLELDSPG
homolog_0488%
EGKSTLQLALEQLKEAKKCLASNGSGLMKAAEALLAAEMVLDRDNYYVTIYCWNGGTVIAIIWFNGQPLEQAEERLARAALKAELDSEG
homolog_0590%
KPKSTLQLALEQLKEAKKLLATNGSGLLKAAEALLAAEMALDRENDYVTIYVWNGGTFIVMIWFNGSPLEQADEKLARAALKLELESG–
homolog_0683%
MGKSTLQYALEQLRQARKLLASSPSGLLVAAEALLAAEMLLERESDYVTIYVWNGGTVFVIIWFNPSPLEQKEEKLARAAVRLELDS––
homolog_0778%
VGKTNLQLAMEQIKEAKKLLASNHEGVAKAAEALLAAEMILDKDNLYVTIYVWKGGTVIVIIWFLGSGLEQAEDKLAVAALQLELDQGP
homolog_0884%
MTQWTLQLALEQAKEAKKALASNGS–LLKAAEALLAAEECLERDN–YVSIYVLNGFSVIVIIWFNGGPLEQAEEKLARAALRLELDSG–
homolog_0969%
–––QTLQLACEQLKEVRKLLATQGNGLLKAALVLLACEMLLDREPYYCTIYVWSGGTVMVIKHFNGSGMEQAEDKLMKAALKLDLDTCN
homolog_1075%
IGKQTLQIALEQLKEAKRLLASS–SILLRAAEALLCMEMLLERENEYVTVYVLSGGQRIVLIFW––SPLELAEEKLARAALKLELDSG–
homolog_1173%
MPVTTINLALDQLKEIRKLLAS–GSGLLSIAEALLVAEALLD––NDFVTIYVWNGPWVIVIIHFTGLPLEKAEEKLAPAALKADLESGG
homolog_1266%
APRSQLSLAIESLKEAKRLLAQNPTGLLKAADALLAAEMCLEK––EYVTIYVW–CPNVILIIFWQGSPLEQADEKLARAALWLEAENGP
homolog_1370%
–GRMTCQLLAEQLKEARKLVVSNPSPLLKAEEALLACEMLLERELDYVTIYVWNPGTVIVIIWYQPGPLEFIEERLAKAALKLELESGP
Consensus
MGKSTLQLALEQLKEAKKLLASNGSGLLKAAEALLAAEMLLDRDNDYVTIYVWNGGTVIVIIWFNGSPLEQAEEKLARAALKLELDSGG
  • Hydrophobic
  • Aromatic
  • Positive
  • Negative
  • Polar

Hover a column to read its composition.

14 sequences, 89 columns, 74 strongly conserved columns.

Structure analysis

A structure is also
a matrix of distances.

Every pair of residues has a distance. Laid out as a matrix, helices show as bands along the diagonal, a hairpin of two strands as a line crossing it, and tertiary contacts as isolated off-diagonal spots. Predicted structures are routinely scored by comparing exactly this kind of matrix.

This one is computed from the packed coordinates of the generic chain in the viewer. Switch to the change view to see which pairs moved when the chain packed.

Hover the matrix to read a pair.

0 Å42 Å and beyond

Simulation and sampling

The landscape is rugged.
The clock is short.

A simulation takes femtosecond steps, and the events that matter take microseconds to days. A walker on a funnelled surface with decoy basins shows why that gap decides what you can conclude. At low temperature it settles in the first well it finds and stays there.

Raise the temperature, or switch on a history-dependent bias that fills visited wells as it goes, the idea behind metadynamics, and it climbs out. This is a Metropolis walker on a constructed two-dimensional surface in reduced units, computed live.

State
Unfolded
Steps
0
Basins visited
1 of 7
Lowest energy reached
0.00
History-dependent bias

Epigenetic age and reprogramming

A clock shift is a reading.
Rejuvenation is a separate claim.

DNA sequence does not age in this sense. What is measured, and what is claimed to be reversed, is the methylation pattern at specific CpG sites, read through a statistical model. That makes the question computational: how a clock is built, what it responds to besides age, and what evidence would separate a real change in biology from a shifted proxy.

A research-methodology demonstration. The cohort below is constructed: 360 synthetic samples and 300 synthetic CpGs generated in your browser from a seeded process, and the clock is fitted in your browser. Nothing here is patient data, a result, a treatment or a product, and Forticia offers none.

Clock

Scenario applied to the test cohort

Change in predicted age
−4.2 years

95% interval −4.4 to −4.1. Chronological age is unchanged.

Held-out error
3.1 years

Mean absolute error on 108 held-out samples, r = 0.982.

CpGs with weight
53 of 300

53% of the weight sits on cell-type-sensitive CpGs.

No age-associated methylation was changed, yet the clock moved by −4.2 years. Everything in that movement comes from the confounders.

Predicted against chronological age, held-out samples. Grey is before, pink is after, joined per sample.
204060800255075100Chronological age (years)Predicted age (years)
The 44 CpGs with the largest clock weights. Pink carries true age signal, blue responds to cell composition, grey carries neither.
+−CpGs ranked by absolute weight
Source of changeAppliedMean change, years95% intervalSpread between samples
Real methylation change0 years-equivalent0.00.0 to 0.00.0
Cell-composition shift0.50−4.2−4.3 to −4.10.6
Batch offset0.000.00.0 to 0.00.0
Technical noise0.300.0−0.1 to +0.10.5
All together−4.2−4.4 to −4.10.8

A real applied shift reads lower than its true size. Regularised regression shrinks coefficients toward zero, so a fitted clock attenuates true effects. This is a known property of such models, and one more reason a clock reading is not a measurement of biological change.

What a clock is

A penalised regression, usually an elastic net, over methylation levels at a few hundred CpG sites, trained to predict an age-related target. Horvath’s 2013 multi-tissue clock uses 353 CpGs and Hannum’s 2013 whole-blood clock uses 71. Second-generation clocks change the target: PhenoAge (513 CpGs) is trained on a phenotypic age measure, GrimAge (1,030 CpGs) on methylation surrogates for smoking and plasma proteins that predict mortality, and DunedinPACE (173 CpGs) on the pace of ageing, with CpGs chosen for test-retest reliability.

What partial reprogramming is

Brief, repeated expression of the Yamanaka factors short of full reprogramming. In 2016, cycles of two days on and five days off extended lifespan in a progeria mouse model; results in normally ageing mice have been more modest. The factors are pro-oncogenic, full reprogramming produced teratomas, and every programme described in the review literature is preclinical.

What is still open

Clocks lose accuracy at older ages and differ between tissues. Technical replicates of the same sample can differ by a median of about one to two and a half years across six widely used clocks, and by 4.5 to 8.6 years at the extreme. Cell-type composition and batch move readings. A falling clock reading and a functional improvement are different claims, and the literature reports cases where they did not track each other.

What would count as evidence

Claim to supportWhat would countFailure it rules out
The reading movedThe shift is larger than replicate noise measured on the same platform, and is reported with an interval.Noise read as effect
It is not the layoutTreated and control samples are randomised across arrays and batches, and batch is in the model.Batch offsets
It is not compositionCell-type fractions are measured or populations are sorted, and the shift survives adjustment.Composition shifts
It is not the clockA clock fixed in advance, plus others trained differently, point the same way. The favourable one is not picked afterwards.Clock shopping
The biology changedOrthogonal readouts move with it, in function, expression and other marks, and persist after the intervention stops.A shifted proxy only
It repeatsAn independent cohort or laboratory reproduces it from the recorded pipeline.One-off or cohort-specific results

Pipelines and provenance

A run is a hash of everything
that went into it.

Every task is identified by its tool version, parameters, environment and the digests of its inputs. Change one thing and only what depends on it is invalidated; the rest is replayed from the store. Pin the environment and the seed, and a repeat reproduces the bytes. Float either, and watch what survives.

A mechanics demonstration: the digests are real SHA-256 values computed in your browser from the parameters and upstream digests shown, and the statistic each task emits is constructed, not a scientific result.

Run explorer0 runs in this session

Run configuration

Aligner sensitivity
Seed
Environment
Task cache

Showing the last run. Change a setting to see which tasks it invalidates.

cached: replayed from the storeexecutedstale: invalidated, will re-execute

Timeline

Log

Run ledger

RunSpecificationExecutedCachedConclusionResult digestTimeCompare

Run comparison

Choose a run as A and another as B in the ledger, or use Run twice without cache.

What we mean by reproducible

Reproducible is not one thing.
It is five different questions.

Each question holds some parts of an experiment fixed and varies one. The dimensions follow the PRIMAD model from the reproducibility literature: platform, research objective, implementation, method, actor and data. Choose a question to see what has to be in the record for it to be answerable.

QuestionWhat variesWhat it tests
NothingDoes the record alone reproduce the result?
ActorIs the record complete without a verbal hand-over?
PlatformDoes the result depend on the hardware it happened to run on?
ImplementationIs the finding a property of the method, or of one piece of code?
DataDoes it generalise, or was it only describing what it had seen?

Run the recorded run again

Expect
Identical outputs for deterministic steps, and a declared tolerance for the rest.
The record must hold
Input digests, code revision, container digest, parameters, seeds and the exact command.

Why bit-identical is the wrong test for some steps

The same 200,000 numbers, summed in different orders, as parallel hardware would. Computed live in your browser in double precision.

Tolerance
Reduction orderResultRelative differenceVerdict
Sequential, reference-13169.0162638744860Reference
Sequential, run again-13169.0162638744860Within tolerance
Reversed order-13169.0162638748462.73e-14Within tolerance
Pairwise tree-13169.0162638748522.78e-14Within tolerance
8 parallel lanes-13169.0162638752265.62e-14Within tolerance
64 parallel lanes-13169.0162638749733.70e-14Within tolerance

Where a tool is not deterministic, the tolerance is declared before the run. A tolerance chosen after seeing the difference is not a tolerance.

Evaluation

A score is only as honest
as the split behind it.

Biological sequences are related by descent, so a random split puts near-copies of training examples in the test set and rewards recall. Clustering by similarity before splitting is the standard remedy. Models pretrained on large sequence corpora add a second route: a test family can be absent from your training set and still be inside the pretraining data.

Each point in the plot is one sequence in an abstract similarity map, grouped into families. Change the split in the first cell and the two measured quantities in the second move with it. Both are computed from the points shown.

split_audit.ipynbPython 3
In [1]
data, corpus = load_similarity_map(families=8)
train, test = random_split(data, test_fraction=0.22)
In [2]
spacing = median(nearest(train, train))
ratio = median(nearest(test, train)) / spacing
exposed = share(nearest(test, corpus) < 1.6 * spacing)
Out [2]
train, test
162, 46
ratio
1.13
exposed
0.57

ratio is how far test points sit from training data, relative to how far training points sit from each other. Near 1 means the test set looks like the training set.

exposed is the share of test points with a pretraining-corpus neighbour inside typical spacing.

Every family appears on both sides. Most test points have a close relative in training.

Out [3]
traintestHover a test point to see its nearest neighbours.

Latest papers

Computational Biology

All papers
  • 5 Oct 2026How many independent proteins are in a benchmark? Error correlation across sequence similarity and the cost to error barsModel errors on related proteins are correlated even at the weakest significant hits. We measure it and what it does to benchmark confidence intervals.21 min read
  • 5 Oct 2026Where identical code gives different answers: a reproducibility budget for a protein-sequence pipelineOne protein pipeline, 192 runs, one factor changed at a time: seeds dominate, thread and BLAS noise is invisible, and one library release flips a verdict.17 min read

RelatedIsolated research workspaces

Further reading

The literature behind the claims on this page.

  1. Accurate structure prediction of biomolecular interactions with AlphaFold 3
  2. CASP16 protein monomer structure prediction assessment
  3. DisProtBench: functional limits of structure prediction in intrinsically disordered regions
  4. De novo design of protein structure and function with RFdiffusion
  5. RFdiffusion exhibits low success rate in de novo design of functional protein binders for biochemical detection
  6. Beware of data leakage from protein LLM pretraining
  7. Current best practices in single-cell RNA-seq analysis: a tutorial
  8. Enhanced sampling in the age of machine learning: algorithms and applications
  9. HH-suite3 for fast remote homology detection and deep protein annotation
  10. Horvath (2013). DNA methylation age of human tissues and cell types. Genome Biology
  11. Hannum et al. (2013). Genome-wide methylation profiles reveal quantitative views of human aging rates. Molecular Cell
  12. Levine et al. (2018). An epigenetic biomarker of aging for lifespan and healthspan (DNAm PhenoAge). Aging
  13. Lu et al. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging
  14. Belsky et al. (2022). DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife
  15. A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking
  16. Partial cellular reprogramming: a deep dive into an emerging rejuvenation technology
  17. Systematically examining reproducibility: a case study for high throughput sequencing using the PRIMAD model and BioCompute Object

Apply for a lab space.

Internal researchers, contractors and academic collaborators can be provisioned without inheriting the rest of Forticia. Writes that change the record wait for a lead.

Request access